# RUUT Labs — full site content > RUUT Labs is a deep-tech company building sovereign agentic AI — the RUUT BRAIN orchestration layer, on-prem EDGEBOX compute, and verified AI agents — for industries that cannot afford to get AI wrong. Source: https://ruutlabs.in. Generated from the same content as the website. --- # RUUT Labs — Intelligence, engineered differently. > RUUT Labs builds sovereign agentic AI: the RUUT BRAIN orchestration layer, on-prem EDGEBOX compute and verified agents, for work that can't get AI wrong. Canonical: https://ruutlabs.in · Last updated: 2026-10-30 ## What is RUUT Labs? RUUT Labs is a deep-tech company building sovereign agentic AI — the RUUT BRAIN orchestration layer, on-prem EDGEBOX compute, and verified AI agents — for industries that cannot afford to get AI wrong. RUUT Labs builds the agentic layer for industries that can't afford to get AI wrong — perceiving the physical world, reasoning across models, verifying every decision, and acting on your own infrastructure. Principles: Model-agnostic · On-premise · Human approval · Edge AI. ## Built for industries that can't afford to get AI wrong - Manufacturing - Real estate - Banking & fintech - Agencies & marketing ## One accountable loop. Real signals. Verifiable action. Every RUUT agent runs the same four-stage loop, so behaviour is predictable and every step can be traced. 1. **Perceive** — VISION and SENSE read the physical world as it happens — cameras, vibration, temperature — with data kept local. (RUUT VISION, RUUT SENSE) 2. **Reason** — RUUT BRAIN interprets signals against your operating rules, routing work across models on compute you control. (RUUT BRAIN) 3. **Verify** — FORGE checks every consequential decision before it executes. Human approval stays in the loop where it matters. (RUUT FORGE) 4. **Act** — Agents carry out the work and report back — FOREMAN, FLOW, VAULT, VECTOR — with a trace of every step. (RUUT FOREMAN, RUUT FLOW, RUUT VAULT, RUUT VECTOR) ## Eight components. One sovereign system. Signals from VISION and SENSE flow into RUUT BRAIN. FORGE verifies every decision before agents act. All of it runs on EDGEBOX. - **RUUT BRAIN** (Orchestration) — The orchestration layer. Routes reasoning across models and coordinates every agent. - **RUUT VISION** (Vision) — Computer vision for physical sites — PPE, fire and intrusion detection, designed for on-prem. - **RUUT SENSE** (Sensors) — Sensor intelligence — vibration, temperature and anomaly signals, fed straight into BRAIN. - **RUUT FORGE** (Verification) — The verification layer. Consequential decisions are checked before they are allowed to act. - **RUUT EDGEBOX** (Sovereign compute) — The on-premise AI appliance that runs the whole RUUT loop inside your own walls. - **RUUT VECTOR** (Embodied AI) — Embodied intelligence — connecting agents to real operational environments. - **RUUT FLOW** (Agencies) — The AI operating system for marketing agencies — one workspace per client brand. - **RUUT VAULT** (Financial institutions) — Agentic infrastructure for banks and financial institutions — credit, collections and compliance. - **RUUT FOREMAN** (Plant safety) — The plant-safety agent. One clear shift-end view of violations, machine signals and observations. ## What runs underneath every RUUT agent. One request path, with verification on both sides of the reasoning. Request path: Request → FORGE input guard → BRAIN graph → Tools + memory → FORGE verify → Action + trace. - **Orchestration.** Agents run as explicit, stateful graphs — planned, checkpointed and resumable, with human approval gates where you need them. - **Governed tools.** Systems are exposed to agents as tools over the Model Context Protocol, each agent with its own allow-list and nothing more. - **Memory.** Durable facts and history in PostgreSQL; documents and past runs in vector retrieval for grounded answers. - **Verification.** Input screening, policy checks on tool calls, output validation and an audit log on every consequential path. - **Observability.** Traces of prompts, retrieval, tool calls, latency and cost per step — with evaluations and replay. *Reference architecture. Third-party names denote technologies RUUT is designed to work with, not partnerships.* ## Runs inside your walls. Stays under your control. RUUT EDGEBOX is the on-prem foundation for the whole stack — for teams that need control over infrastructure, data movement and operational continuity. - **Local data control.** Architecture designed so data stays inside your environment. - **Your models.** Run open-weight models on hardware you control, or route to providers you approve. - **Human approval.** Approval gates you can require for any consequential action. - **Full audit trail.** Every signal, decision and action is logged and reviewable. ## Agentic AI, built around your business. Beyond our own products, RUUT Labs designs and ships custom agentic systems — from one focused workflow to a full sovereign deployment. Process: Scope (The problem, the data, the risk.) → Prototype (A working agent on real inputs.) → Verify (Checks, approvals and evaluation.) → Deploy (On-prem, cloud or hybrid.). - **Custom agent design.** Agents and architecture built around RUUT BRAIN for your workflow. - **Integration.** Connects to the systems you already operate — no rip-and-replace. - **Deployment your way.** On-premise, cloud or hybrid, set by your requirements. ## Frequently asked questions ### What is RUUT Labs? RUUT Labs is a deep-tech company building sovereign agentic AI — the RUUT BRAIN orchestration layer, on-prem EDGEBOX compute, and verified AI agents — for industries that cannot afford to get AI wrong. Alongside its own products, RUUT Labs designs and builds custom agentic systems for organisations. ### What is RUUT BRAIN? RUUT BRAIN is the orchestration layer of the RUUT system. It interprets signals from VISION and SENSE against your operating rules, plans work as explicit, resumable graphs, and routes reasoning across models. Before any consequential action, BRAIN hands the decision to RUUT FORGE for verification. ### What is agentic AI, and how is it different from a chatbot? Agentic AI is software that pursues a goal by planning steps, using tools and taking actions, rather than only replying to messages. A chatbot answers; an agent acts. That is why RUUT pairs agents with verification and human approval: when software can act, every consequential step needs to be checked and traceable. ### Can RUUT run on-premise? Yes, the RUUT system is designed to run on-premise — on RUUT EDGEBOX or on infrastructure you already operate — so data can stay inside your environment. Whether a deployment also runs without an internet connection depends on the components and models chosen, and is scoped with each project. ### What is RUUT FORGE, and how does it verify AI decisions? RUUT FORGE is the verification layer every consequential decision passes through before it acts. It screens inputs and retrieved content for manipulation, checks proposed tool calls against policy, validates outputs, and routes decisions that need a person to human approval. Every check is logged so it can be reviewed later. ### Which industries is RUUT designed for? RUUT is designed for industries where AI mistakes are costly: manufacturing and plant operations, banking and financial services, real estate, and marketing agencies. The same loop — perceive, reason, verify, act — adapts to each through domain agents such as FOREMAN, VAULT and FLOW. ### Does RUUT Labs build custom AI agents? Yes. Beyond its own products, RUUT Labs designs and builds custom agentic systems — from one focused workflow to a full sovereign deployment. Projects move through four stages: scope, prototype, verify and deploy, and integrate with the systems you already run rather than replacing them. ### Which AI models does RUUT work with? RUUT is model-agnostic. RUUT BRAIN routes work across models, including open-weight models you host yourself and commercial models where your policies allow them. Model choice is made per task and per deployment, so you are not locked into a single provider. ### Where is RUUT Labs based? RUUT Labs is based in India and is an operating brand of QZERO LABS (OPC) Pvt. Ltd. Because the RUUT system is designed to run on infrastructure you control, the team works with organisations wherever their data and operations are. The fastest way to reach the team is contact@ruutlabs.in. ### How do I start a project, and what does it cost? Start by telling us what you want to solve through the contact form or ATLAS. RUUT Labs does not publish pricing: every deployment is scoped to your data, infrastructure and risk tolerance. The first conversation is about the problem, the constraints and whether an agentic system is the right answer. --- # The RUUT system > The RUUT system: eight components in one sovereign loop — VISION and SENSE perceive, BRAIN reasons, FORGE verifies, agents act, all on EDGEBOX. Canonical: https://ruutlabs.in/platform · Last updated: 2026-10-30 ## Overview The RUUT system is one loop built from eight components. VISION and SENSE perceive the physical world, RUUT BRAIN reasons across models, FORGE verifies every consequential decision, and agents act — all of it able to run on EDGEBOX, inside your walls. ## Principles - **Verification before action.** Nothing consequential executes without passing FORGE: input screening, policy checks, output validation and human approval where you require it. - **Sovereign by design.** Every component is designed to run on-premise or in your private cloud, with models you choose. - **Agents that touch the physical world.** Perception from cameras and sensors, and action through equipment — not only chat. - **One accountable loop.** Perceive, reason, verify, act — the same path for every agent, with a full trace of every step. ## Components by layer - **RUUT BRAIN** — Reason · Orchestration: The orchestration layer. Routes reasoning across models and coordinates every agent. - **RUUT VISION** — Perceive · Vision: Computer vision for physical sites — PPE, fire and intrusion detection, designed for on-prem. - **RUUT SENSE** — Perceive · Sensors: Sensor intelligence — vibration, temperature and anomaly signals, fed straight into BRAIN. - **RUUT FORGE** — Verify · Verification: The verification layer. Consequential decisions are checked before they are allowed to act. - **RUUT EDGEBOX** — Compute · Sovereign compute: The on-premise AI appliance that runs the whole RUUT loop inside your own walls. - **RUUT VECTOR** — Act · Embodied AI: Embodied intelligence — connecting agents to real operational environments. - **RUUT FLOW** — Act · Agencies: The AI operating system for marketing agencies — one workspace per client brand. - **RUUT VAULT** — Act · Financial institutions: Agentic infrastructure for banks and financial institutions — credit, collections and compliance. - **RUUT FOREMAN** — Act · Plant safety: The plant-safety agent. One clear shift-end view of violations, machine signals and observations. --- # RUUT BRAIN — the orchestration layer > RUUT BRAIN is the orchestration layer of the RUUT system: it plans work as resumable graphs, routes reasoning across models and hands decisions to FORGE. Canonical: https://ruutlabs.in/platform/brain · Last updated: 2026-10-01 ## What is RUUT BRAIN? RUUT BRAIN is the orchestration layer of the RUUT system. It interprets signals from VISION and SENSE against your operating rules, plans work as explicit, resumable graphs, routes reasoning across models, and hands every consequential decision to FORGE before agents act. ## What RUUT BRAIN does - **Plans work as explicit graphs.** Every task becomes a stateful graph of steps — planned, checkpointed and resumable — so long-running work survives restarts and every branch can be inspected. - **Routes reasoning across models.** BRAIN chooses a model per step: an open-weight model on your hardware for routine work, a larger model where policy allows. You are never locked to one provider. - **Coordinates every layer.** It turns VISION and SENSE observations into context, calls governed tools, consults memory, and passes consequential decisions to FORGE before any agent acts. ## Where it sits in the loop Layer: Reason. BRAIN is the Reason stage of the RUUT Agent Loop. It receives signals from VISION and SENSE (Perceive), sends proposed actions to FORGE (Verify), and directs agents such as FOREMAN, FLOW and VAULT (Act). It runs on EDGEBOX or on infrastructure you already operate. ## How it is deployed BRAIN is designed to run on-premise on RUUT EDGEBOX, in your private cloud, or as a hybrid where sensitive steps stay local. Model endpoints, tools and approval rules are configured per deployment. ## Typical uses - Interpreting plant signals against safety rules before alerting a supervisor - Running a multi-step back-office workflow that pauses for human approval - Answering operational questions from your own documents with retrieval ## Questions about RUUT BRAIN ### Is RUUT BRAIN a large language model? No. RUUT BRAIN is an orchestration layer, not a model. It decides which model, tool or agent handles each step, keeps state across steps, and routes consequential decisions to FORGE. The models it uses can be open-weight models you host or providers you approve. ### Can RUUT BRAIN work with our existing systems? Yes. Existing systems are connected to BRAIN as governed tools over the Model Context Protocol, each with an explicit allow-list. That means BRAIN can read from and act on the systems you already run, without replacing them, and only within the permissions you set. ### What happens if a step fails midway? Because work runs as a checkpointed graph, a failed or interrupted step can be retried or resumed from the last checkpoint instead of starting over. Every step, retry and approval is recorded in the trace, so the path to any outcome can be reviewed. ## Related - RUUT FORGE - RUUT VISION - RUUT SENSE - RUUT EDGEBOX --- # RUUT VISION — computer vision for physical sites > RUUT VISION reads camera feeds to detect events such as missing PPE, fire or intrusion and sends observations to RUUT BRAIN. Designed for on-prem. Canonical: https://ruutlabs.in/platform/vision · Last updated: 2026-10-01 ## What is RUUT VISION? RUUT VISION is the computer-vision layer of the RUUT system. It reads camera feeds from plants, sites and facilities to detect events such as missing PPE, fire or intrusion, and passes structured observations to RUUT BRAIN. It is designed to run on-premise. ## What RUUT VISION does - **Detects safety and security events.** VISION is designed to recognise events such as missing personal protective equipment, fire or smoke, and intrusion into restricted zones from standard camera feeds. - **Turns frames into observations.** Instead of raw video, VISION emits structured observations — what, where, when, confidence — that RUUT BRAIN can reason over alongside rules and sensor data. - **Keeps footage local.** Inference is designed to run close to the cameras, on EDGEBOX or local hardware, so footage does not need to leave the site. ## Where it sits in the loop Layer: Perceive. VISION is part of the Perceive stage of the RUUT Agent Loop, alongside SENSE. Its observations flow into RUUT BRAIN, which decides whether they matter under your rules; any resulting action passes through FORGE before an agent such as FOREMAN carries it out. ## How it is deployed VISION is designed to run on-premise, next to existing cameras, on RUUT EDGEBOX or comparable local compute. Zones, event types and alert rules are configured per site. ## Typical uses - PPE compliance checks in defined zones of a plant - Early fire or smoke observations for a safety team - Restricted-area intrusion alerts outside working hours ## Questions about RUUT VISION ### Does RUUT VISION need new cameras? RUUT VISION is designed to work with standard IP camera feeds, so many sites can start with the cameras they already have. Camera placement, resolution and lighting affect what can be detected reliably, which is why every deployment begins with a site assessment. ### Does video leave our premises? VISION is designed so inference runs on-premise, close to the cameras, and only structured observations move on to RUUT BRAIN. Whether any footage is retained, and for how long, is a configuration decision made with you for each deployment. ### Can VISION act on its own? No. VISION perceives; it does not act. Its observations go to RUUT BRAIN, and any consequential action — stopping a line, escalating an alert — is checked by FORGE and can require human approval before an agent carries it out. ## Related - RUUT SENSE - RUUT BRAIN - RUUT FOREMAN - RUUT EDGEBOX --- # RUUT SENSE — sensor intelligence > RUUT SENSE turns machine and environment signals — vibration, temperature and similar telemetry — into anomaly observations for RUUT BRAIN. Canonical: https://ruutlabs.in/platform/sense · Last updated: 2026-10-01 ## What is RUUT SENSE? RUUT SENSE is the sensor-intelligence layer of the RUUT system. It turns machine and environment signals — vibration, temperature and similar telemetry — into anomaly observations that RUUT BRAIN can reason over alongside vision and operating context. ## What RUUT SENSE does - **Reads machine and environment signals.** SENSE ingests telemetry such as vibration and temperature from equipment and environments, through the interfaces your sites already use. - **Flags anomalies, not just thresholds.** It is designed to compare signals against each asset's normal behaviour, so unusual patterns can be raised as observations before they become incidents. - **Adds context for reasoning.** Observations carry the asset, location and time, so RUUT BRAIN can combine them with vision events and operating rules into one picture. ## Where it sits in the loop Layer: Perceive. SENSE is part of the Perceive stage of the RUUT Agent Loop. Together with VISION it gives RUUT BRAIN a view of the physical world; decisions that follow are verified by FORGE before agents act. ## How it is deployed SENSE is designed to run on-premise on RUUT EDGEBOX or local gateways, connecting to existing sensors and industrial interfaces. Signal mappings and anomaly rules are set up per site. ## Typical uses - Early vibration anomalies on rotating equipment - Temperature drift in process or storage areas - Combining machine signals with safety observations in a shift report ## Questions about RUUT SENSE ### What kind of sensors can RUUT SENSE use? SENSE is designed for common machine and environment telemetry such as vibration and temperature, collected through the sensors and industrial interfaces a site already has. The exact signals and integration path are agreed during scoping for each deployment. ### Is RUUT SENSE a predictive-maintenance product? SENSE is the sensing layer: it turns raw signals into anomaly observations. What happens next — a maintenance ticket, an alert, a line in the shift report — is decided by RUUT BRAIN under your rules and verified by FORGE, so SENSE can support maintenance workflows without acting alone. ### Can SENSE run without cloud connectivity? SENSE is designed to run on-premise, close to the equipment. Whether a full deployment operates without any internet connection depends on the other components and models chosen, and is decided during scoping. ## Related - RUUT VISION - RUUT BRAIN - RUUT FOREMAN - RUUT EDGEBOX --- # RUUT FORGE — the verification layer > RUUT FORGE checks every consequential AI decision against readable, versioned rules before it acts, returns a clear verdict and records why. Canonical: https://ruutlabs.in/platform/forge · Last updated: 2026-10-30 ## What is RUUT FORGE? RUUT FORGE is the verification layer of the RUUT system. It checks every consequential AI decision against versioned, readable rules before it executes, returns a clear verdict — CERTIFIED, REJECTED or UNCERTAIN — routes what needs a person to human approval, and records why. ## What RUUT FORGE does - **Screens what goes in.** FORGE checks user input and retrieved content for manipulation such as prompt injection, and for data that should not reach a model, before reasoning begins. - **Checks every action against readable rules.** Each proposed tool call or decision is checked against versioned rule packs — which agent, which system, which action, which limits — written so compliance and safety teams can read and approve them. - **Gives a verdict, not a guess.** Every check ends CERTIFIED, REJECTED or UNCERTAIN, with its reasons and the evidence used. UNCERTAIN goes to a person, so nothing consequential proceeds on a guess — and every verdict is recorded. ## Where it sits in the loop Layer: Verify. FORGE is the Verify stage of the RUUT Agent Loop and sits on both sides of the reasoning: it guards what enters RUUT BRAIN and checks what BRAIN proposes before any agent acts. Nothing consequential reaches the Act stage without a FORGE verdict. ## How it is deployed FORGE is deployed wherever BRAIN runs — on-premise on EDGEBOX, in a private cloud, or hybrid — and can also be called through its API to check decisions made outside the RUUT system. Rule packs, approval rules and escalation paths are defined with your team. ## Typical uses - Requiring supervisor approval before an agent stops a production line - Blocking an agent from calling a system outside its allow-list - Showing an auditor, months later, which rules approved a decision and why ## Questions about RUUT FORGE ### What does it mean to verify an AI decision? Verifying an AI decision means checking it before it executes: is the input trustworthy, is the proposed action allowed by policy, is the output supported by evidence, and does a person need to approve it. RUUT FORGE runs these checks on every consequential path and records each one. ### What does a FORGE verdict say? Every checked decision ends in one of three verdicts: CERTIFIED, when it passes the rules that apply; REJECTED, when a rule fails, with the reason; or UNCERTAIN, when the evidence is not enough to decide, so a person reviews it. FORGE never calls an AI output true or correct — it records which rules were applied, to which evidence. ### How is FORGE different from AI guardrails? Most guardrails filter text: they block unsafe words in a prompt or a reply. FORGE governs decisions and actions. It checks what an agent is about to do against versioned rules your team has approved, decides whether a person must confirm it, and keeps a record that shows afterwards exactly why an action was allowed or refused. ### Who writes the rules FORGE applies? Rules are written in a readable, versioned rule language and grouped into domain packs — for example, an industrial-safety pack for plants or a pack for a lending workflow. Your compliance or safety team can read, review and approve exactly what the system enforces, and every verdict records the version of the rules it was checked against. ### Does FORGE replace human approval? No. FORGE decides when human approval is needed and routes the decision to the right person; it does not remove people from consequential decisions. You choose which actions always require approval and which can proceed automatically within policy. ### Can FORGE stop prompt injection? FORGE screens inputs and retrieved content for prompt-injection patterns and restricts what any agent is allowed to do, which limits the impact even if an attack is not detected. No single control is perfect, so FORGE combines screening, least-privilege tools, output checks and approvals. ### Can FORGE be used without the rest of RUUT? Yes. FORGE is designed as standalone verification infrastructure with an API, so it can check decisions made by agents and models outside the RUUT system as well as inside it. Within RUUT, every consequential step proposed by RUUT BRAIN passes through FORGE automatically, with no extra integration work. ## Related - RUUT BRAIN - RUUT VAULT - RUUT FOREMAN - RUUT EDGEBOX --- # RUUT EDGEBOX — the on-premise AI appliance > RUUT EDGEBOX is the on-premise AI appliance of the RUUT system: perception, open-weight models, retrieval and FORGE verification inside your facility. Canonical: https://ruutlabs.in/platform/edgebox · Last updated: 2026-10-30 ## What is RUUT EDGEBOX? RUUT EDGEBOX is the on-premise AI appliance of the RUUT system. It runs perception, open-weight models, retrieval and FORGE verification inside your own facility, so sensitive data stays on site and operations keep running without depending on an external connection. ## What RUUT EDGEBOX does - **The whole loop in one box.** EDGEBOX hosts VISION, SENSE, BRAIN and FORGE together on your premises: cameras and sensors in, verified actions out, with nothing required to leave the site. - **Your models, run locally.** Open-weight models run on the appliance's own accelerator, with a local knowledge store for retrieval over your documents. External providers are used only for steps your policy explicitly allows. - **Built to keep running.** Critical inference stays local, so perception, reasoning and verification continue through network outages. Updates are staged and applied only with your approval, and every change is recorded. ## Where it sits in the loop Layer: Compute. EDGEBOX is not a stage of the loop — it is the substrate beneath all four. Perceive, Reason, Verify and Act all run on EDGEBOX, which is why it sits under every RUUT deployment that has to stay on-premise, from a plant floor to a bank's controlled environment. ## How it is deployed EDGEBOX is installed at your site and sized to the workload: the number of cameras and sensors, the models to run and the agents in use. Installation, model setup and the FORGE rule packs for the site are part of every project scope. ## Typical uses - A plant that needs camera footage and safety records to stay on site - A financial institution that must keep customer data inside a controlled environment - Sites where connectivity is limited, intermittent or not allowed ## Questions about RUUT EDGEBOX ### Is RUUT EDGEBOX a hardware product? Yes. RUUT EDGEBOX is an on-premise AI appliance: accelerator hardware with the RUUT runtime, models and local storage, installed at your site. Its configuration is sized during scoping to the cameras, sensors, models and agents a deployment needs, and the same software can also run on infrastructure you already operate. ### What data leaves the site with EDGEBOX? By design, none that you have not approved. Camera footage, sensor readings, documents and the records FORGE keeps are processed and stored on the appliance. If a deployment uses an external model for a particular step, that route is set in policy, limited to that step, and recorded every time it is used. ### Why run AI on-premise instead of in the cloud? For control, continuity and predictable cost. Sensitive or regulated data stays inside your walls; operations keep running when connectivity does not; and inference on hardware you own does not grow with every API call. Cloud still has its place, and RUUT can route specific steps to it where your policy allows. ### Do we need EDGEBOX to use RUUT? No. EDGEBOX is for deployments that need to run on-premise. The RUUT system can also run on infrastructure you already operate, in a private cloud, or in a hybrid arrangement — the choice depends on your data, latency and control requirements. ## Related - RUUT BRAIN - RUUT VISION - RUUT FORGE - RUUT SENSE --- # RUUT VECTOR — embodied intelligence > RUUT VECTOR connects verified agent decisions to physical operational environments, so real-world action follows the same perceive, reason, verify loop. Canonical: https://ruutlabs.in/platform/vector · Last updated: 2026-10-01 ## What is RUUT VECTOR? RUUT VECTOR is the embodied-intelligence layer of the RUUT system. It connects verified agent decisions to physical operational environments — equipment, machines and site systems — so that action in the real world follows the same perceive, reason, verify loop. ## What RUUT VECTOR does - **Connects decisions to the physical world.** VECTOR is the bridge from a verified decision to equipment, machines and site systems — the point where software meets operations. - **Acts only after verification.** Every physical action passes through FORGE first, with human approval where your policy requires it. - **Reports what actually happened.** Actions are confirmed back into the trace, so the record shows not just what was decided but what was carried out. ## Where it sits in the loop Layer: Act. VECTOR is part of the Act stage of the RUUT Agent Loop. It receives decisions that RUUT BRAIN proposed and FORGE verified, and carries them into operational environments, reporting outcomes back into the trace. ## How it is deployed VECTOR is designed for on-premise deployment alongside EDGEBOX, integrated with the specific equipment and control systems in scope. Integrations are defined case by case. ## Typical uses - Carrying out an approved operational action on site equipment - Coordinating a physical workflow step after a verified decision - Closing the loop between an alert and the action taken ## Questions about RUUT VECTOR ### What is embodied AI? Embodied AI is AI that acts in the physical world through machines and equipment, not only through text or screens. In the RUUT system, RUUT VECTOR is the layer that connects verified agent decisions to those physical environments. ### Is RUUT VECTOR a robot? No. RUUT VECTOR is the software layer that connects agent decisions to operational equipment and systems. What it controls depends on the integrations agreed for a deployment, and every action is checked by FORGE first. ### How is physical action kept safe? Every action VECTOR carries out has first been proposed by RUUT BRAIN and verified by FORGE against policy, with human approval required wherever you decide it should be. Each step is recorded in the trace for review. ## Related - RUUT BRAIN - RUUT FORGE - RUUT EDGEBOX - RUUT FOREMAN --- # RUUT FLOW — the AI operating system for agencies > RUUT FLOW is an AI operating system for digital marketing agencies: one workspace per client brand for content, scheduling, CRM and approvals. Canonical: https://ruutlabs.in/platform/flow · Last updated: 2026-10-01 ## What is RUUT FLOW? RUUT FLOW is an AI operating system for digital marketing agencies. Each client brand gets its own workspace for brand-grounded content, scheduling, CRM and approvals, so agency teams can run multi-client work from one place. Website: https://flow.ruutlabs.in ## What RUUT FLOW does - **One workspace per client brand.** Each client brand gets its own workspace, so content and context for one client never mix with another's. - **Brand-grounded content.** Drafts are grounded in each brand's own material, so AI-assisted content starts from the client's voice rather than a generic one. - **Scheduling, CRM and approvals.** Planning, client relationships and approval steps live in the same place, so work moves from draft to approved to scheduled without switching tools. ## Where it sits in the loop Layer: Act. FLOW is an Act-stage application of the RUUT system for agencies. It applies the same principles as the rest of the loop — grounded generation, approval before anything is published, and a record of each step. ## How it is deployed RUUT FLOW is available as a hosted application at flow.ruutlabs.in; agencies sign in and set up one workspace per client brand. ## Typical uses - Running content for several client brands from one team - Keeping client approvals in one place before anything is scheduled - Grounding AI drafts in each brand's own material ## Questions about RUUT FLOW ### Who is RUUT FLOW for? RUUT FLOW is built for digital marketing agencies that manage several client brands. It gives each brand its own workspace for grounded content, scheduling, CRM and approvals, so an agency team can run multi-client work from one place. ### Does RUUT FLOW publish content automatically? RUUT FLOW keeps approvals in the workflow, so content moves through the approval steps your agency and clients define before it is scheduled. Nothing needs to go out without the sign-off you require. ### Where can I see RUUT FLOW? RUUT FLOW is available at flow.ruutlabs.in. For a walkthrough with the team, use the contact form or ask ATLAS to set up an introduction. ## Related - RUUT BRAIN - RUUT FORGE - RUUT VAULT --- # RUUT VAULT — agentic infrastructure for financial institutions > RUUT VAULT is agentic infrastructure for banks and financial institutions: credit, collections and compliance, where scorecards decide and FORGE verifies. Canonical: https://ruutlabs.in/platform/vault · Last updated: 2026-10-30 ## What is RUUT VAULT? RUUT VAULT is agentic infrastructure for banks and financial institutions, built on the RUUT system. It runs credit, collections and compliance workflows in which language models handle the conversation, approved scorecards make the decision, and FORGE verifies every consequential step before it takes effect. ## What RUUT VAULT does - **Credit decisions with reasons.** Decisions come from the scorecards and policies your credit team approves — never from a language model — and every outcome carries reason codes a customer, auditor or regulator can read. - **Compliance written as rules.** Lending and conduct rules are written as readable FORGE rule packs. A disbursement, limit change or collection step runs only after the rules that apply have passed, and the check is recorded. - **Agents that follow conduct rules.** Onboarding, servicing, collections and grievance agents speak the customer's language and follow your conduct policy: costs shown plainly, no pressure tactics, and a person on hand for disputes. ## Where it sits in the loop Layer: Act. VAULT is an Act-stage application of the RUUT system for financial services. RUUT BRAIN coordinates each workflow, approved scorecards decide, FORGE verifies every consequential step against your rule packs, and VAULT carries it out with a full record. ## How it is deployed VAULT is designed to run inside the institution's controlled environment — on-premise on EDGEBOX or in a private cloud — and to connect to existing core banking, lending and payment systems. The regulated activity and the loan book stay with the institution. ## Typical uses - Credit decisions for thin-file customers from consented income and cash-flow data - Collections outreach that follows conduct rules and records every contact - A reviewable record of every disbursement, limit change and decision reason ## Questions about RUUT VAULT ### What does RUUT VAULT do? RUUT VAULT is agentic infrastructure for banks, lenders and other financial institutions. It runs credit, collections and compliance workflows on the RUUT system: language models handle conversations and paperwork, approved scorecards make the decisions, and RUUT FORGE checks every consequential step — a disbursement, a limit change, a collection action — before it happens, with a full record. ### Does an AI model make the credit decision? No. In RUUT VAULT, language models never decide credit. They help collect information, explain terms and draft communication. The decision itself comes from scorecards and policies your credit team approves, carries readable reason codes, and is verified by FORGE against your rules before anything is disbursed. ### Is RUUT VAULT a bank or lender? No. RUUT VAULT is infrastructure software. The regulated activity — lending, holding the loan book, collecting — stays with the bank or financial institution using it. VAULT is designed to make those workflows faster, consistent with the institution's own policies, and fully reviewable by its risk, compliance and audit teams. ## Related - RUUT FORGE - RUUT BRAIN - RUUT EDGEBOX --- # RUUT FOREMAN — the plant-safety agent > RUUT FOREMAN is a plant-safety agent: it combines vision and sensor observations, reasons over safety rules, verifies actions and reports each shift. Canonical: https://ruutlabs.in/agents/foreman · Last updated: 2026-10-30 ## What is RUUT FOREMAN? RUUT FOREMAN is a plant-safety agent built on the RUUT system. It combines VISION and SENSE observations, reasons over site safety rules with RUUT BRAIN, verifies proposed actions with FORGE, and reports violations, machine signals and safety observations at the end of each shift. ## What RUUT FOREMAN does - **Watches the floor.** FOREMAN combines VISION observations such as missing PPE with SENSE signals such as vibration anomalies, by zone and time. - **Reasons over your safety rules.** RUUT BRAIN matches observations to the site's own rules, so an alert reflects your procedures rather than a generic threshold. - **Reports every shift.** At the end of each shift, FOREMAN compiles violations, machine signals and safety observations into one report for the people responsible. ## Where it sits in the loop Layer: Act. FOREMAN is the worked example of the full RUUT Agent Loop: VISION and SENSE perceive, RUUT BRAIN reasons over safety rules, FORGE verifies any action and routes approvals, and FOREMAN acts — alerting, escalating and reporting — with a trace of every step. ## How it is deployed FOREMAN is designed to run on-premise on RUUT EDGEBOX with the site's cameras and sensors. Zones, rules, alert recipients and report formats are configured with the site's safety team. ## Typical uses - Shift-end safety reports for plant managers - Supervisor alerts when a rule is breached in a defined zone - Combining safety observations with machine signals in one view ## Questions about RUUT FOREMAN ### What does RUUT FOREMAN report? At the end of each shift, RUUT FOREMAN compiles one report covering safety violations observed by VISION, machine signals from SENSE and other safety observations, organised so the people responsible can review what happened and what was done about it. ### Can FOREMAN stop a machine? Any action beyond alerting is proposed by RUUT BRAIN and checked by FORGE; stopping equipment is the kind of action you would normally require a supervisor to approve. What FOREMAN may do, and with whose approval, is set with your safety team. ### Does FOREMAN replace safety staff? No. FOREMAN gives safety staff a continuous view of the floor and a clear shift report. Decisions about people and procedures stay with your team; FOREMAN surfaces what it observed and routes consequential actions for approval. ### How are a site's safety rules set up in FOREMAN? The site's rules — zones, PPE requirements, machine thresholds and which actions need approval — are written with your safety team as a FORGE industrial-safety rule pack. The rules stay readable and versioned, so every alert and every report shows which rule was matched and which version of the rules was in force at the time. ## Related - RUUT VISION - RUUT SENSE - RUUT BRAIN - RUUT FORGE --- # Solutions > Industries RUUT is designed for. Canonical: https://ruutlabs.in/solutions · Last updated: 2026-10-30 ## Industries - **Manufacturing** — On a plant floor, AI mistakes have physical consequences. RUUT combines camera and sensor perception with reasoning over your own safety and operating rules, verifies every consequential action before it happens, and runs on-premise so footage and machine data stay on site. - **Real estate** — Property transactions involve large sums, many parties and long timelines. RUUT brings agentic workflows to documents, milestones and approvals, with every consequential step — a payment release, a record change, a commitment to a customer — verified by FORGE and recorded for review. - **Banking & fintech** — Financial services need AI that can be governed. Through RUUT VAULT, banks and financial institutions run credit, collections and compliance workflows with agents that have least-privilege access, decisions that come from approved scorecards, FORGE checks on every consequential step, human approval where you decide, and a full trace — on infrastructure you control. - **Agencies & marketing** — Agencies run many brands at once, each with its own voice, calendar and approval chain. RUUT FLOW gives every client brand its own AI workspace for brand-grounded content, scheduling, CRM and approvals, so teams scale output without mixing brands or skipping sign-off. --- # Agentic AI for manufacturing and plant operations > How RUUT applies agentic AI to plants: VISION and SENSE perceive the floor, BRAIN reasons over safety rules, FORGE verifies, FOREMAN reports each shift. Canonical: https://ruutlabs.in/solutions/manufacturing · Last updated: 2026-10-30 ## Summary On a plant floor, AI mistakes have physical consequences. RUUT combines camera and sensor perception with reasoning over your own safety and operating rules, verifies every consequential action before it happens, and runs on-premise so footage and machine data stay on site. ## What makes AI hard here - **Safety observations arrive too late.** Violations and near-misses are often found after the shift, from manual checks. Continuous perception can surface them while they still matter. - **Signals live in separate systems.** Camera events, machine telemetry and safety procedures rarely meet in one place. Reasoning across them needs a layer that sees all three. - **Automation must not act blindly.** Stopping a line or escalating an incident is consequential. Any automated step needs checks and, where it matters, a person's approval. ## The RUUT approach - **Perceive:** VISION reads camera feeds for missing PPE, fire or intrusion; SENSE reads vibration and temperature. - **Reason:** RUUT BRAIN matches observations to the site's own safety and operating rules. - **Verify:** FORGE checks every proposed action and routes stop-line decisions to a supervisor. - **Act:** FOREMAN alerts the right people and compiles a report at the end of every shift. ## Components - RUUT FOREMAN - RUUT VISION - RUUT SENSE - RUUT EDGEBOX ## Common questions ### How can AI improve plant safety without adding risk? By keeping AI in the roles it does well — continuous observation and fast correlation — and keeping consequential actions behind verification and human approval. In the RUUT system, VISION and SENSE observe, BRAIN reasons over your rules, and FORGE checks any action before FOREMAN carries it out. ### Does plant data have to leave the site? No. The RUUT system is designed to run on-premise on RUUT EDGEBOX, close to the cameras and sensors, so footage and machine data can stay on site. Which steps, if any, use external services is decided with you during scoping. --- # Agentic AI for real estate > How RUUT applies verified agentic AI to real estate: documents, milestones and approvals, with FORGE checks and human approval on consequential steps. Canonical: https://ruutlabs.in/solutions/real-estate · Last updated: 2026-10-30 ## Summary Property transactions involve large sums, many parties and long timelines. RUUT brings agentic workflows to documents, milestones and approvals, with every consequential step — a payment release, a record change, a commitment to a customer — verified by FORGE and recorded for review. ## What makes AI hard here - **High-value steps need certainty.** A single wrong release or record change is costly. Automation is only useful if each consequential step is checked before it happens. - **Many parties, one record.** Buyers, developers, lenders and agents each hold part of the picture. A shared, traceable record reduces disputes and rework. - **Long timelines lose context.** A transaction runs across many milestones and documents. Each step needs the context of the ones before it. ## The RUUT approach - **Perceive:** Documents, milestones and enquiries enter the system as structured inputs. - **Reason:** RUUT BRAIN prepares each step — a credit check, a release, a response — under your policies. - **Verify:** FORGE checks every consequential step against policy and evidence, with human approval where required. - **Act:** Agents carry out verified steps — notices, updates, releases — and record each one for review. ## Components - RUUT FORGE - RUUT BRAIN - RUUT EDGEBOX ## Common questions ### Can AI safely automate escrow releases? Escrow releases can be prepared and checked automatically, but the release itself should pass verification and, where your policy says so, human approval. In the RUUT system, each release is tied to a milestone, verified by FORGE against policy and evidence, and recorded, so automation speeds the work without removing control. ### How does RUUT keep a property transaction traceable? Every step is recorded: which document or milestone triggered it, what RUUT BRAIN prepared, which FORGE rules it was checked against, who approved it and what happened. Disputes can then be settled against one shared history instead of scattered emails and files, and each new step starts with the context of the ones before it. --- # Agentic AI for banking and financial services > How RUUT applies verified agentic AI in financial services: governed tools, policy checks, human approval and a full audit trail, deployable on-prem. Canonical: https://ruutlabs.in/solutions/banking · Last updated: 2026-10-30 ## Summary Financial services need AI that can be governed. Through RUUT VAULT, banks and financial institutions run credit, collections and compliance workflows with agents that have least-privilege access, decisions that come from approved scorecards, FORGE checks on every consequential step, human approval where you decide, and a full trace — on infrastructure you control. ## What makes AI hard here - **Every decision must be explainable.** Regulators, auditors and customers expect to know why a decision was made. Automation needs a trace, not just an outcome. - **Data must stay controlled.** Customer and transaction data cannot travel freely. AI has to run where policy allows it to run. - **Agents need narrow permissions.** An agent that can call any system is a risk. Each one should have exactly the access its task requires. ## The RUUT approach - **Perceive:** Applications, documents and requests arrive as structured inputs, screened by FORGE. - **Reason:** RUUT BRAIN plans each workflow as a checkpointed graph, using only allow-listed tools. - **Verify:** FORGE enforces policy on every tool call and routes defined decisions for human approval. - **Act:** VAULT and other agents complete the workflow, and every step is written to an audit log. ## Components - RUUT VAULT - RUUT FORGE - RUUT BRAIN - RUUT EDGEBOX ## Common questions ### How do you make AI agents auditable in financial services? By recording every step: the input, what was retrieved, which tool was called with which permissions, which policy checks ran, who approved what, and what happened. The RUUT system writes this trace for every run, and RUUT FORGE ensures no consequential step skips its checks. ### Can RUUT run inside our own infrastructure? Yes. The RUUT system is designed to run on-premise or in a private cloud, so data stays within environments you control. Model choice is part of the same decision: open-weight models can run locally, and external providers are used only where your policy allows. --- # Agentic AI for marketing agencies > How RUUT applies agentic AI to agencies: RUUT FLOW gives each client brand a workspace for grounded content, scheduling, CRM and approvals. Canonical: https://ruutlabs.in/solutions/agencies · Last updated: 2026-10-30 ## Summary Agencies run many brands at once, each with its own voice, calendar and approval chain. RUUT FLOW gives every client brand its own AI workspace for brand-grounded content, scheduling, CRM and approvals, so teams scale output without mixing brands or skipping sign-off. ## What makes AI hard here - **Brand voice drifts at scale.** Generic AI drafts sound the same for every client. Content has to start from each brand's own material. - **Work is spread across tools.** Drafts, calendars, client threads and approvals usually live in different places, which slows every handoff. - **Nothing should publish unapproved.** Client trust depends on sign-off. Automation has to respect the approval chain, not route around it. ## The RUUT approach - **Perceive:** Each brand's material, briefs and client requests are gathered into its own workspace. - **Reason:** Drafts and plans are generated grounded in that brand's material only. - **Verify:** Approval steps defined by the agency and client gate everything before scheduling. - **Act:** FLOW schedules approved work and keeps the client record up to date. ## Components - RUUT FLOW - RUUT BRAIN - RUUT FORGE ## Common questions ### What is an AI operating system for agencies? An AI operating system for agencies brings content creation, scheduling, client relationship management and approvals into one place, with AI assistance grounded in each client's brand. RUUT FLOW is built this way: one workspace per client brand, so context never mixes between clients. ### How do agencies keep AI content on-brand? By grounding generation in each brand's own material and keeping a human approval step before anything is published. RUUT FLOW gives every brand its own workspace for this, so drafts start from the client's voice and still pass the sign-off the agency and client agree. --- # Custom agentic AI builds > RUUT Labs designs and builds custom agentic AI systems — scope, prototype, verify, deploy — integrated with your systems, on-prem, cloud or hybrid. Canonical: https://ruutlabs.in/services · Last updated: 2026-10-30 ## Overview Beyond our own products, RUUT Labs designs and ships custom agentic systems — from one focused workflow to a full sovereign deployment — built on the same loop and the same verification layer. ## Process 1. **Scope.** We start with the problem, the data and the risk: what the agent should do, what it must never do, what systems it touches and who approves what. The output is a written scope, not a slide. 2. **Prototype.** A working agent on real inputs, early. Prototypes run on your data in a controlled environment so everyone can see actual behaviour before decisions are made. 3. **Verify.** We define the checks — policies, approval gates, evaluation sets — and run them. An agent moves forward only when its behaviour on your cases is understood and acceptable. 4. **Deploy.** On-premise, cloud or hybrid, integrated with the systems you already run. Deployment includes monitoring, traces and a clear path for changes. ## Engagements - **Custom agent design.** Agents and architecture built around RUUT BRAIN for a specific workflow, with FORGE verification on every consequential step. - **Integration.** Your existing systems connected as governed tools — each agent with an explicit allow-list — so nothing has to be ripped out. - **Deployment your way.** On-premise on RUUT EDGEBOX, in your private cloud, or hybrid, set by your data, latency and control requirements. ## Before you start ### What does a custom agentic AI project with RUUT Labs look like? It moves through four stages: scope, prototype, verify and deploy. Scope defines what the agent must and must never do; the prototype runs on real inputs early; verification defines and runs the checks and approvals; deployment integrates with your systems on-premise, in the cloud or hybrid. ### Do we have to use RUUT products in a custom build? Custom builds are designed around RUUT BRAIN for orchestration and RUUT FORGE for verification, because those give every agent the same accountable loop. Other components — VISION, SENSE, EDGEBOX — are used only where the problem needs them. ### How is a custom project priced? RUUT Labs does not publish pricing. Each project is scoped to your data, infrastructure and risk tolerance, and the first conversation is about the problem and whether an agentic system is the right answer at all. --- # Sovereign AI, on-premise > Sovereign AI with RUUT: run perception, reasoning and verification on-prem on EDGEBOX or a private cloud, with models you choose and a full audit trail. Canonical: https://ruutlabs.in/sovereignty · Last updated: 2026-10-30 ## Overview Sovereign AI means the organisation that depends on an AI system controls where it runs, which models it uses, what data it sees and who approves what it does. The RUUT system is designed for that from the start. ## What sovereign AI means in practice Sovereignty is not a single feature. It is four decisions that stay with you: **where** the system runs, **which models** it uses, **what data** it can see, and **who approves** consequential actions. The RUUT system keeps each of those decisions configurable per deployment. | Decision | How RUUT handles it | | --- | --- | | Where it runs | On-premise on RUUT EDGEBOX, on infrastructure you already operate, in a private cloud, or hybrid. | | Which models | Open-weight models hosted locally; external providers only for steps where your policy allows them. | | What data | Agents reach systems only through governed tools with explicit allow-lists; inputs are screened by FORGE. | | Who approves | Human approval gates you can require for any consequential action, enforced by FORGE. | *Sovereignty decisions and where they are enforced in the RUUT system.* ## RUUT EDGEBOX [RUUT EDGEBOX](/platform/edgebox) is the on-premise compute foundation for the whole stack. It is designed to host perception, orchestration and verification inside your facility, so camera footage, sensor data and operational records can be processed where they are produced. ## Connectivity Keeping critical inference local reduces dependence on external connectivity for the steps that matter most. Whether a particular deployment runs entirely without an internet connection depends on the components and models chosen, and is confirmed during scoping — we would rather tell you exactly than promise in general. ## Accountability Every signal, decision and action is written to a trace you can review: what came in, what was retrieved, which tool was called with which permissions, which checks ran, who approved, and what happened. That record stays with you. ## Sovereign AI, explained ### What is sovereign AI? Sovereign AI is AI whose operating organisation controls where it runs, which models it uses, what data it can access and who approves its consequential actions. In practice that usually means on-premise or private-cloud deployment, model choice, governed data access and enforced approval gates — the four decisions the RUUT system keeps configurable. ### What is the difference between on-premise AI and edge AI? On-premise AI runs on infrastructure inside your organisation's own facilities. Edge AI runs close to where data is produced — next to cameras, sensors or machines — often on-premise as well. RUUT EDGEBOX is designed for both: on your premises, close to the signals it processes. ### Can we use cloud models and still keep control? Yes, selectively. RUUT BRAIN can route specific steps to approved external providers while keeping sensitive steps on local models, and FORGE screens what is sent. Which steps may use which providers is a policy you set, not a default. --- # Developers > How RUUT integrates: governed tools over MCP, model-agnostic orchestration, and an agent-readable site with llms.txt, Markdown twins and an MCP server. Canonical: https://ruutlabs.in/developers · Last updated: 2026-10-30 ## Overview RUUT is designed to be integrated, not just used. Your systems become governed tools for agents, models are interchangeable, and even this website can be queried by your own AI agent. ## Integration model - **Governed tools.** Systems are exposed to agents as tools over the [Model Context Protocol](https://modelcontextprotocol.io), each agent with an explicit allow-list and nothing more. - **Model-agnostic orchestration.** RUUT BRAIN routes each step to a model: open-weight models you host, or providers you approve. - **Verification on every path.** RUUT FORGE screens inputs, checks tool calls against policy and validates outputs before anything executes. - **Traces by default.** Every run records inputs, retrieval, tool calls, checks, approvals, latency and cost per step. ## This website is agent-readable We built ruutlabs.in the way we build agent systems: content first, machine-readable by design. Every page is server-rendered HTML, and the same content is available in formats agents prefer. | Interface | What it gives you | | --- | --- | | [/llms.txt](/llms.txt) | A curated index of the site for language models. | | [/llms-full.txt](/llms-full.txt) | The full site content as one plain-text document. | | `/{page}.md` | A clean Markdown version of any page, e.g. [/platform/forge.md](/platform/forge.md). | | [/.well-known/ruutlabs.json](/.well-known/ruutlabs.json) | Company facts and the product list as JSON. | | `/api/mcp` | A public, read-only MCP server your AI agent can query directly. | ## Connect your agent to RUUT Labs over MCP The RUUT Labs MCP server exposes read-only tools — `search_site`, `get_product`, `list_solutions`, `get_company_facts` — and one write tool, `request_intro`, which sends a contact request only after you confirm it. Add it to any MCP client that supports remote servers over Streamable HTTP: ```json { "mcpServers": { "ruutlabs": { "type": "http", "url": "https://ruutlabs.in/api/mcp" } } } ``` > **Reference architecture.** Third-party names on this page denote technologies RUUT is designed to work with, not partnerships or endorsements. --- # Insights > Technical writing from RUUT Labs. Canonical: https://ruutlabs.in/insights · Last updated: 2026-10-30 ## Articles - [What is an agentic operating system?](/insights/what-is-an-agentic-operating-system) — An agentic operating system runs AI agents like an OS runs programs: it schedules work, keeps state, governs tools, checks actions and records every step. - [Sovereign AI on-premise: when your data can't leave the building](/insights/sovereign-ai-on-premise) — Sovereign AI: you control where AI runs, which models it uses, what data it sees and who approves its actions. When on-prem is right, and what it takes. - [Verifying AI decisions before they act: the case for a verification layer](/insights/verifying-ai-decisions-before-they-act) — A verification layer checks AI decisions before they take effect: input screening, policy checks, output validation and human approval, all recorded. --- # What is an agentic operating system? > An agentic operating system runs AI agents like an OS runs programs: it schedules work, keeps state, governs tools, checks actions and records every step. Canonical: https://ruutlabs.in/insights/what-is-an-agentic-operating-system · Last updated: 2026-10-01 ## TL;DR By RUUT Labs. Published 2026-10-01. - An agentic operating system is the layer that runs AI agents: it schedules their work, keeps their state, controls their tools, verifies consequential actions and records what happened. - It exists because agents act. Once software can change records or move equipment, the hard problems are permissions, verification and accountability — not text generation. - RUUT organises this layer around one loop: Perceive, Reason, Verify, Act. ## Introduction **An agentic operating system is the software layer that runs AI agents the way an operating system runs programs.** It decides what each agent works on, keeps track of where every task is, controls which tools and data each agent may use, checks consequential actions before they happen, and records everything. The models inside it generate plans and language; the operating system makes the whole arrangement dependable. The term is useful because it moves attention away from the model and toward the system around it. That shift matters most for organisations that want agents to do real work — update a record, release a payment, alert a supervisor — rather than draft text for a person to copy. ## From chatbots to agents A chatbot answers. An agent acts. An **AI agent** is a program that uses a model to decide on and carry out actions toward a goal: it plans a step, calls a tool, observes the result and decides what to do next. Research such as *ReAct* showed how interleaving reasoning with actions lets a model use tools to complete multi-step tasks, and that pattern is now the backbone of most agent designs. The difference sounds small and is not. When the output is text, a mistake is something a person reads and discards. When the output is an action, a mistake is something that has already happened. Every question that follows — who may do what, how decisions are checked, how anyone can find out what went wrong — comes from that one change. ## Why agents need an operating system A single agent in a demo can get by with a prompt and a few tools. A set of agents running inside a business cannot. They need the same things programs needed before operating systems existed: a way to share resources safely, to recover from failure, to be isolated from each other, and to be observed. | Classic operating system | Agentic operating system | | --- | --- | | Schedules processes | Plans and schedules agent tasks as explicit, resumable steps | | Manages memory | Manages state, conversation history and retrieved knowledge | | Controls access to devices and files | Controls access to tools and data through allow-lists | | Enforces permissions | Verifies consequential actions and requires approval where set | | Logs system events | Records a trace of every input, decision, check and action | *The analogy is imperfect, but each classic responsibility has an agentic counterpart.* ## What an agentic operating system does ### Orchestration Work is planned as a graph of steps rather than one long prompt. Each step has inputs, outputs and a checkpoint, so long-running work can pause for an approval, survive a restart and resume where it stopped. Orchestration also decides which model handles which step — a small local model for routine classification, a larger one where the task and policy allow. ### Governed tools Agents reach the outside world only through tools: an API, a database query, a device command. An agentic operating system exposes those tools through a standard interface — the [Model Context Protocol](https://modelcontextprotocol.io) is one open option — and gives each agent an explicit allow-list. An agent that was never granted a tool cannot call it, however it is prompted. ### Memory and grounding Agents need durable facts and relevant documents. **Retrieval-augmented generation** fetches trusted passages before a model answers, so claims can be traced to sources. The operating system decides what each agent may retrieve and keeps durable state separate from a single conversation. ### Verification Before anything consequential executes, the proposed action is checked: is the input trustworthy, is the action within policy, is the output supported by evidence, does a person need to approve it? This is the part most often missing from agent prototypes, and the part that decides whether an agent can be trusted with real work. We cover it in depth in [Verifying AI decisions before they act](/insights/verifying-ai-decisions-before-they-act). ### Observability Every run leaves a trace: inputs, retrieved context, model calls, tool calls, checks, approvals, outcomes and timing. Traces are how teams debug agents, evaluate changes and answer the question an auditor will eventually ask — *why did the system do that?* ## The RUUT Agent Loop RUUT Labs organises its agentic operating system around one loop that every agent follows, whether it is watching a plant floor or preparing a credit decision: 1. **Perceive** — VISION and SENSE turn cameras and sensors into structured observations. 2. **Reason** — [RUUT BRAIN](/platform/brain) interprets observations against operating rules and plans the next step across models. 3. **Verify** — [RUUT FORGE](/platform/forge) checks every consequential decision and routes approvals to people. 4. **Act** — agents such as [RUUT FOREMAN](/agents/foreman) carry out the verified action and report back, with a trace of every step. Using one loop for every agent is a deliberate constraint. It makes behaviour predictable, because every agent passes the same checkpoints, and it makes accountability simple, because every action can be traced back through verification and reasoning to the signal that started it. ## How to evaluate an agentic platform If you are assessing any platform for running agents — ours or anyone else's — these questions separate a demo from a system: - Can each agent be limited to an explicit list of tools and actions? - Which actions are checked before they execute, and how is the policy defined? - Can you require human approval for specific actions, and is the approval recorded? - Can long-running work pause, survive a failure and resume? - Can you see a complete trace of any run after the fact? - Can it run where your data must stay, with the models you choose? Frameworks such as the NIST [AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) are a good companion to these questions: they frame the governance an organisation needs around any AI system, and an agentic operating system is where much of that governance is actually enforced. ## The short version > A model decides what to say. An agentic operating system decides what an agent is allowed to do, checks it before it happens, and remembers that it did. ## References - [ReAct: Synergizing Reasoning and Acting in Language Models (Yao et al.)](https://arxiv.org/abs/2210.03629) — arXiv - [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al.)](https://arxiv.org/abs/2005.11401) — arXiv - [Model Context Protocol — Specification](https://modelcontextprotocol.io/specification) — Model Context Protocol - [AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) — National Institute of Standards and Technology (NIST) --- # Sovereign AI on-premise: when your data can't leave the building > Sovereign AI: you control where AI runs, which models it uses, what data it sees and who approves its actions. When on-prem is right, and what it takes. Canonical: https://ruutlabs.in/insights/sovereign-ai-on-premise · Last updated: 2026-10-01 ## TL;DR By RUUT Labs. Published 2026-10-01. - Sovereign AI is AI whose operator controls four things: where it runs, which models it uses, what data it can reach, and who approves its consequential actions. - On-premise deployment is the strongest form of the first control, but sovereignty also depends on model choice, data governance and approvals. - On-premise is right when data, latency or continuity demand it; it is a trade-off, not a virtue in itself. ## Introduction **Sovereign AI is AI whose operating organisation controls where it runs, which models it uses, what data it can access and who approves its consequential actions.** For many organisations the first of those controls — where it runs — is the one that decides everything else, because some data simply cannot leave the building. This article explains what sovereignty means in practice, when running AI on your own premises is the right choice, and what it takes to do well. It is written to be useful whatever platform you choose. ## Sovereignty is four decisions, not one It is common to equate sovereign AI with on-premise hosting. Hosting matters, but an AI system running on your hardware can still send data to an external model, give an agent broader access than it needs, or act without anyone's approval. Real control comes from four decisions staying with you: | Decision | The question it answers | | --- | --- | | Location | Where does processing physically happen, and under whose control? | | Models | Which models are used for which tasks, and can they be changed? | | Data access | What can each agent read or change, and through which interfaces? | | Approval | Which actions need a person's sign-off, and is that enforced? | ## When on-premise is the right answer Running AI on your own premises has real costs: hardware, capacity planning, operations and updates. It is worth it when one or more of these is true: - **The data is sensitive or regulated.** Camera footage of workers, customer financial records and confidential documents often carry contractual or legal limits on where they may be processed. - **Latency matters.** A safety alert on a plant floor is more useful in a second than in a minute; processing next to the cameras avoids a round trip. - **Connectivity is unreliable.** Sites with limited or intermittent connections need critical inference to keep working locally. - **Volume makes transfer impractical.** Continuous video and high-frequency sensor data are expensive to move; processing at the edge sends only observations onward. If none of these apply, a private cloud or a hybrid design may give you the controls you need with less to operate. The honest answer depends on the workload, and we would rather say so than default to hardware. ## On-premise, edge and hybrid **On-premise AI** runs on infrastructure inside your own facilities. **Edge AI** runs close to where data is produced — beside cameras, sensors or machines. The two often coincide: a compute unit on a factory floor is both. **Hybrid** designs keep sensitive steps local and send specific, approved steps to external models, with what may be sent decided by policy rather than convenience. ## Models you can actually choose Sovereignty over models has two parts. First, the ability to run capable open-weight models on your own hardware for the tasks that must stay local. Second, an orchestration layer that is model-agnostic, so each step can use the most suitable model and switching providers does not mean rebuilding the system. In the RUUT system, [RUUT BRAIN](/platform/brain) routes each step to a model under policy, and local models run on [RUUT EDGEBOX](/platform/edgebox). ## Governed data access Keeping data on-premise is not enough if every agent can read everything. Agents should reach systems only through governed tools with explicit allow-lists, and inputs — including retrieved documents — should be screened for manipulation before they reach a model. The [OWASP Top 10 for LLM Applications](https://genai.owasp.org/llm-top-10/) is a practical checklist of the risks this governance has to address. ## Approval and accountability The last control is the one most often left implicit: who approves what. A sovereign system makes approval rules explicit and enforced — for example, an agent may alert a supervisor on its own, but stopping a line requires the supervisor's approval. Every approval, and every action, is recorded in a trace the organisation keeps. ## What running on-premise actually involves On-premise AI moves responsibilities from a provider to you. Plan for them before the hardware arrives, not after: - **Capacity.** Size compute for the real workload — the number of camera streams, sensor rates, the models in use and how many agents run at once — with headroom for growth. - **Model lifecycle.** Open-weight models improve quickly. Decide how new versions are evaluated against your own cases before they replace the old ones. - **Patching and security.** On-premise systems need the same patching, access control and monitoring as any other critical server. - **Observability.** Traces and health metrics must be collected locally, and someone must be responsible for reading them. - **Skills.** Someone on your side, or a partner, has to own the system day to day. ## Hybrid in practice Most real deployments are not purely local. A common pattern keeps perception and anything touching personal or operational data on-premise, uses local models for routine reasoning, and allows a specific, low-sensitivity step — summarising a public document, say — to use an external model. What makes it sovereign is not where every step runs, but that **the boundary is a written policy the system enforces**, and every crossing is recorded. ## A practical checklist 1. List the data each AI workflow touches and where each item is allowed to be processed. 2. Decide which steps must use local models and which may use external providers. 3. Define, per agent, the tools and actions it may use — and nothing else. 4. Write down which actions require human approval, and by whom. 5. Require a complete trace of every run, stored where you control it. 6. Plan capacity, updates and support for on-premise components before go-live. > **How RUUT approaches it.** The RUUT system is designed so each of the four decisions stays configurable per deployment: EDGEBOX for location, BRAIN for model routing, governed tools for data access, and FORGE for approvals and verification. Read more on [Sovereignty](/sovereignty). ## References - [AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) — National Institute of Standards and Technology (NIST) - [OWASP Top 10 for Large Language Model Applications](https://genai.owasp.org/llm-top-10/) — OWASP Gen AI Security Project - [Model Context Protocol — Specification](https://modelcontextprotocol.io/specification) — Model Context Protocol --- # Verifying AI decisions before they act: the case for a verification layer > A verification layer checks AI decisions before they take effect: input screening, policy checks, output validation and human approval, all recorded. Canonical: https://ruutlabs.in/insights/verifying-ai-decisions-before-they-act · Last updated: 2026-10-01 ## TL;DR By RUUT Labs. Published 2026-10-01. - An AI verification layer checks decisions before they take effect: it screens inputs, enforces policy on proposed actions, validates outputs and routes some decisions to people. - It is needed because agents act on text they did not write, and no model or filter is reliable on its own. - Verification works as layers — screening, least privilege, policy, validation, approval, trace — so one failure does not become an incident. ## Introduction **An AI verification layer is the part of an agentic system that checks a decision before it takes effect.** It screens what goes into the model, enforces policy on what an agent proposes to do, validates the output against rules and evidence, routes decisions that need a person to human approval, and records every check. Nothing consequential reaches the real world without passing it. This article makes the case for treating verification as a dedicated layer rather than a few prompts and filters, and describes what that layer needs to contain. ## Why agents need verification Three properties of agents make verification necessary rather than optional. - **Agents act.** Their output is a change in a system — a record, a payment, a machine state — not a paragraph someone can ignore. - **Agents read text they did not write.** Documents, emails, web pages and tool results all flow into the model, and any of them can carry instructions. Research on *indirect prompt injection* has shown how content retrieved by an application can steer the model that reads it. - **Models are probabilistic.** The same input can produce different outputs, and a fluent answer is not necessarily a correct one. None of these can be fixed by a better prompt. They are fixed by architecture: checks that sit outside the model and do not depend on the model behaving well. ## What a verification layer checks | Check | When | Purpose | | --- | --- | --- | | Input screening | Before reasoning | Detect manipulation and data that should not reach a model | | Tool policy | Before every tool call | Allow only the actions this agent may take, within limits | | Output validation | Before an output is used | Confirm it is supported by evidence and fits the rules | | Human approval | Before defined actions | Put a person's decision on the actions that warrant it | | Audit trace | Always | Record every input, check, approval and outcome | ### Input screening User input and retrieved content are checked for instruction-like patterns, attempts to override the system's rules, and data that policy says must not be sent to a particular model. Screening reduces risk; it does not remove it, which is why the following layers exist. ### Least privilege and tool policy Each agent has an explicit allow-list of tools, and each tool call is checked against policy: which agent, which system, which action, which limits. If an agent is manipulated into wanting something it was never granted, the call fails. **Least privilege limits the blast radius of every other failure.** ### Output validation Before an output is used, it is checked against rules and evidence. A grounded answer should be traceable to the passages it came from; a proposed value should fall within expected bounds; a structured result should match its schema. Outputs that fail are replaced with a safe fallback or escalated. ### Human approval Good systems decide in advance which actions need a person, route them to the right person with the context needed to decide, and record the decision. Asking a human to approve everything is not oversight; it is a queue people learn to click through. ### The audit trace Every check, approval and outcome is written to a trace. The trace is how you debug, how you evaluate a change before rolling it out, and how you answer an auditor, a customer or your own team when they ask why something happened. ## Layers, not a single gate No single control is reliable enough on its own. Screening misses things; models can be confidently wrong; people approve under time pressure. Verification works because the layers are independent: an injection that slips past screening still meets least-privilege tools; an action within the allow-list still meets policy limits; a consequential action still meets approval; and everything is recorded. The [OWASP Top 10 for LLM Applications](https://genai.owasp.org/llm-top-10/) catalogues the failure modes these layers are designed to contain. ## Verification in the RUUT system In the RUUT Agent Loop, verification is its own stage. [RUUT FORGE](/platform/forge) sits on both sides of [RUUT BRAIN](/platform/brain): it guards what enters reasoning and checks what reasoning proposes before any agent acts. For example, when [RUUT FOREMAN](/agents/foreman) observes a safety violation, an alert to the shift supervisor can proceed within policy, but stopping the line is routed to the supervisor for approval — and both paths are recorded. This website applies the same idea at small scale. ATLAS, our site concierge, screens each question, answers only from published content, checks its answer against what it retrieved, and hands off to a person when it cannot ground a reply. The agent trace you see in the ATLAS panel is that pipeline, step by step. ## Common objections ### "Won't verification slow the agent down?" Most checks are fast, deterministic code — allow-lists, schema validation, policy rules — and add little time compared with a model call. The checks that take real time, such as human approval, are reserved for actions where speed matters less than being right. Fast where it can be, careful where it must be. ### "Isn't human approval a bottleneck?" Only if everything needs it. The point of a policy is to decide in advance which actions are routine and which are consequential. Routine actions proceed within limits; consequential ones go to the right person with the context to decide quickly. Approval stays rare enough to be taken seriously. ### "Won't a better model make this unnecessary?" Better models make fewer mistakes, but agents still read untrusted text and still act in systems where one error is costly. Verification does not assume the model is weak; it assumes that any component can fail and that the consequences of action deserve an independent check. ## Questions to ask of any AI system that acts 1. What is checked before a consequential action executes, and who defines the rules? 2. What is each agent prevented from doing, regardless of its instructions? 3. Which actions require a person, and is that enforced by the system rather than by convention? 4. How are retrieved documents and tool results screened before the model reads them? 5. Can you reconstruct exactly why a given action happened? ## References - [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection (Greshake et al.)](https://arxiv.org/abs/2302.12173) — arXiv - [OWASP Top 10 for Large Language Model Applications](https://genai.owasp.org/llm-top-10/) — OWASP Gen AI Security Project - [AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) — National Institute of Standards and Technology (NIST) - [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al.)](https://arxiv.org/abs/2005.11401) — arXiv --- # Glossary of agentic AI terms > Plain definitions of agentic AI terms. Canonical: https://ruutlabs.in/glossary · Last updated: 2026-10-30 ## Agentic AI Agentic AI is software that pursues a goal by planning steps, choosing tools and taking actions, rather than only generating a reply. An agentic system decides what to do next, observes the result and continues. Because it acts, it needs verification, permissions and a record of what it did. ## AI agent An AI agent is a program that uses a model to decide on and carry out actions toward a goal — calling tools, reading data, writing records — within the permissions it has been given. Agents differ from chatbots in that their output is an action in a system, not only text. ## Agentic operating system An agentic operating system is the layer that runs AI agents the way an operating system runs programs: it schedules their work, manages state and memory, controls which tools each may use, verifies consequential actions and records everything. RUUT BRAIN and FORGE together form this layer in the RUUT system. ## RUUT Agent Loop The RUUT Agent Loop is the four-stage path every RUUT agent follows: Perceive (VISION, SENSE), Reason (RUUT BRAIN), Verify (FORGE) and Act (agents such as FOREMAN). Using one loop for every agent makes behaviour predictable and makes every step traceable, from signal to action. ## AI verification layer An AI verification layer checks AI decisions before they take effect: it screens inputs, enforces policy on proposed actions, validates outputs against evidence and rules, and routes decisions that need a person to human approval. In the RUUT system this layer is RUUT FORGE. ## Human-in-the-loop Human-in-the-loop describes an AI system in which people review or approve certain decisions before they take effect. Well-designed systems decide in advance which actions need approval, route them to the right person, and record who approved what, instead of asking a human to check everything. ## Sovereign AI Sovereign AI is AI whose operating organisation controls where it runs, which models it uses, what data it can access and who approves its consequential actions. It usually implies on-premise or private-cloud deployment, model choice and enforced governance rather than dependence on one external provider. ## On-premise AI On-premise AI runs on computing infrastructure inside an organisation's own facilities rather than in a third-party cloud. It keeps data within a controlled environment and can reduce dependence on external connectivity, at the cost of owning capacity planning, hardware and operations. ## Edge AI Edge AI runs models close to where data is produced — beside cameras, sensors or machines — instead of sending raw data to a distant data centre. It is used where latency, bandwidth, privacy or connectivity make central processing impractical, such as plant floors and remote sites. ## On-prem LLM An on-prem LLM is a large language model hosted on infrastructure an organisation controls, typically an open-weight model. It lets sensitive text and documents be processed without leaving that environment. Its capability, hardware cost and maintenance must be weighed against hosted alternatives for each task. ## Model-agnostic A model-agnostic system can use different AI models interchangeably — open-weight or commercial, small or large — without being rebuilt. It lets each task use the most suitable model and avoids lock-in to a single provider. RUUT BRAIN routes work across models in this way. ## Model Context Protocol (MCP) The Model Context Protocol is an open protocol for connecting AI applications to external tools and data sources through a standard interface. A system exposes tools over MCP once, and any compatible AI client can use them, subject to the permissions the server enforces. ## Retrieval-augmented generation (RAG) Retrieval-augmented generation is a technique in which a system first retrieves relevant passages from a trusted source and then has a model answer using those passages. It grounds answers in specific documents and lets each claim be traced to a source, reducing unsupported statements. ## Grounding Grounding means constraining an AI system's output to information from an identified, trusted source — documents, records or sensor data — instead of the model's general knowledge. A grounded answer can be checked against its source; an ungrounded one can only be trusted or doubted. ## Prompt injection Prompt injection is an attack in which text supplied to an AI system — by a user or hidden in retrieved content — tries to override the system's instructions. Defences combine input screening, least-privilege tool access, output checks and approval for consequential actions, because no single filter is complete. ## Least-privilege tool access Least-privilege tool access gives each AI agent only the tools and permissions its task requires, through an explicit allow-list. It limits the damage a wrong decision or a manipulated input can cause, because an agent cannot call systems or actions it was never granted. ## Agent trace An agent trace is the step-by-step record of an AI agent's run: its inputs, retrieved context, model calls, tool calls, checks, approvals and outcomes, with timing. Traces make agent behaviour reviewable after the fact and are the basis for debugging, evaluation and audit. ## PPE detection PPE detection uses computer vision to check whether people in a defined area are wearing required personal protective equipment, such as helmets or high-visibility vests. Results depend on camera placement, lighting and rules per zone, so deployments are configured and validated site by site. --- # About RUUT Labs > About RUUT Labs: a deep-tech company building sovereign agentic AI — the RUUT BRAIN orchestration layer, on-prem EDGEBOX compute and verified AI agents. Canonical: https://ruutlabs.in/company · Last updated: 2026-10-30 ## Who we are RUUT Labs is a deep-tech company building sovereign agentic AI — the RUUT BRAIN orchestration layer, on-prem EDGEBOX compute, and verified AI agents — for industries that cannot afford to get AI wrong. ## Mission Make agentic AI trustworthy enough to run in the physical world — not just in a chat window. ## How we work - **Verify before acting.** When software can act, every consequential step should be checked first and recorded always. - **Control stays with the operator.** The organisation that depends on an AI system should decide where it runs, which models it uses and who approves what it does. - **Say less, show the system.** We describe what we build precisely and let the architecture speak. No inflated claims, no borrowed logos. ## Team and careers The team works across applied machine learning, systems engineering, verification and full-stack product engineering, and builds for real operational environments. We are interested in engineers who want their work running in real operations, not demos. If that is you, write to us with what you have built. ## Company details - Brand: RUUT Labs - Founder: Ajeet Kumar Shukla founded RUUT Labs. - Operated by: QZERO LABS (OPC) Pvt. Ltd. - Based in: India - Email: contact@ruutlabs.in - Tagline: Intelligence, engineered differently. RUUT Labs is an operating brand of QZERO LABS (OPC) Pvt. Ltd. --- # Contact RUUT Labs > Talk to RUUT Labs about a custom agentic AI build, an on-prem deployment or a product demo. No public pricing — every deployment is scoped. Canonical: https://ruutlabs.in/contact · Last updated: 2026-10-30 ## Tell us what you're building. Share the problem and the constraints. The team replies personally — usually with questions before answers. No public pricing; every deployment is scoped to your data, infrastructure and risk tolerance. ## How to reach the team - Email: contact@ruutlabs.in - Contact form: https://ruutlabs.in/contact - Company: RUUT Labs, a brand of QZERO LABS (OPC) Pvt. Ltd., India --- # Privacy notice > What RUUT Labs collects and why. Canonical: https://ruutlabs.in/privacy · Last updated: 2026-10-01 ## Summary This notice explains what QZERO LABS (OPC) Pvt. Ltd. ("RUUT Labs", "we") collects through ruutlabs.in, why, and what you can ask us to do with it. We keep collection to what we need to reply to you and to run the site safely. ## What we collect | Data | When | Why | | --- | --- | --- | | Name, work email, company, interest and message | When you send the contact form, or ask ATLAS or our MCP server to request an introduction | To reply to you and discuss your enquiry | | ATLAS conversations | When you use ATLAS, the site assistant | To answer your question and improve answers. Kept for 30 days, then deleted | | ATLAS usage events (no message text) | When you use ATLAS | To monitor reliability, speed and cost | | A one-way hash of your IP address | When you use the contact form, ATLAS or the MCP server | To limit abuse (rate limiting). We do not store your IP address itself | | Aggregate, cookieless analytics | When you visit pages | To understand which pages are useful. No cookies and no cross-site tracking | Before an ATLAS conversation is stored or sent to a language model, email addresses and phone numbers in it are removed automatically. Please don't share sensitive personal information with ATLAS. ## Cookies We do not use advertising or tracking cookies. Your browser stores your theme preference (light or dark) locally on your device; it is not sent to us. ## Who processes data for us - **Vercel** — hosts the website and provides cookieless analytics. - **Neon** — stores contact enquiries and ATLAS records in a database. - **Upstash** — stores short-lived rate-limiting counters. - **Resend** — sends the email notification of your enquiry to us and the confirmation to you. - **Language-model providers (OpenRouter and Google)** — generate ATLAS answers. We use free service tiers, under whose terms providers may use submitted prompts to improve their services; this is why personal details are removed before sending. ## How long we keep it - Contact enquiries: as long as needed to respond and maintain our business relationship, then deleted on request or when no longer needed. - ATLAS conversations: 30 days. - Usage events and rate-limit data: only as long as needed for reliability and abuse prevention. ## Your choices You can ask us to access, correct or delete personal data we hold about you, or to stop using it, by writing to [contact@ruutlabs.in](mailto:contact@ruutlabs.in). We will respond within a reasonable time and as required by applicable law. ## Changes If we change how we handle data, we will update this page and its date. --- # Terms of use > Terms for using ruutlabs.in. Canonical: https://ruutlabs.in/terms · Last updated: 2026-10-01 ## Summary These terms apply to your use of ruutlabs.in, operated by QZERO LABS (OPC) Pvt. Ltd.. By using the site you agree to them. ## Information on this site Content on this site describes our products, services and approach. It is provided for general information, may change, and is not an offer or a contractual commitment. Any engagement is governed by a separate written agreement. ## ATLAS, the site assistant ATLAS answers questions using content published on this site. It can be wrong, and it hands off to a person when it cannot ground an answer. ATLAS answers are not professional advice and do not create any commitment from RUUT Labs. ## Machine-readable interfaces We publish llms.txt, Markdown versions of pages, a JSON facts file and a read-only MCP server so that people and their AI tools can read this site. Please use them reasonably; automated access is rate-limited, and abusive use may be blocked. ## Intellectual property The RUUT name, logos, product names and site content belong to RUUT Labs. You may quote or link to our content with attribution. Third-party names appear only to describe technologies we work with and do not imply partnership or endorsement. ## Liability The site is provided as is. To the extent permitted by law, we are not liable for losses arising from use of the site or reliance on its content. ## Governing law and contact These terms are governed by the laws of India. Questions: [contact@ruutlabs.in](mailto:contact@ruutlabs.in).