# agnte > agnte is the control layer for your AI workforce: coding agents run as one team on your machines and subscriptions, with budgets and memory that never compacts. Full text of every page: https://agnte.ai/llms-full.txt ## Main pages - [agnte | The Agent Orchestration Platform for AI Coding Agents](https://agnte.ai/): agnte is the control layer for your AI workforce: coding agents run as one team on your machines and subscriptions, with budgets and memory that never compacts. - [Case Study: How agnte Found and Fixed Its Own Blind Spot in a Day](https://agnte.ai/case-study): 1,035 commits in four days, an MVP marked done that a new user could not sign up to, and nine root causes fixed the same day. The Idp story. - [Contact agnte | agnte](https://agnte.ai/contact): Contact the agnte team about sales, enterprise needs or any question. Tell us about your team and we will reply by email. - [agnte Crew: 36 AI Agent Roles Orchestrated From Idea to Release](https://agnte.ai/crew): Meet agnte Crew, the AI engineering team behind every project: business analyst, architect, developer, QA, security, SRE and more, orchestrated from idea to release. - [agnte Pricing: Starter, Professional and Enterprise Plans](https://agnte.ai/pricing): agnte plans for every stage: Starter, Professional and Enterprise. Every plan includes the full pipeline, model routing and budget controls. - [Privacy Policy | agnte](https://agnte.ai/privacy): What personal data agnte collects through its website, desktop app licensing and payments, why, who processes it, how long it is kept and your rights. - [agnte Platform: Agent Orchestration for AI Coding Agents](https://agnte.ai/product): Explore every agnte capability: surfaces like the Web Console and MCP server, automations like code review and QA, and the platform behind fleets, budgets and memory. - [Refund Policy | agnte](https://agnte.ai/refund-policy): agnte refund policy: a full refund within 14 days of your first payment, cancellation at any time, and how refunds are processed by Paddle, our reseller. - [agnte Security: Sandboxed AI Agents, Central Policy, Audit-Ready Logs](https://agnte.ai/security): agnte runs AI coding agents in sandboxes on your own machines with one central policy, scoped access, encrypted secrets and an audit journal of every action. - [Terms of Service | agnte](https://agnte.ai/terms): The terms for using agnte and its website: licenses and subscriptions, payments through Paddle, acceptable use, your data, liability and termination. - [Why agnte: Agent Orchestration vs Single-Session Coding Agents](https://agnte.ai/why-agnte): Coding agents are great at one task. A product is thousands. See how agnte adds memory, a multi-machine fleet, budgets, 36 roles and a closed test loop. ## Platform - [Findings Ledger: Every Issue Routed to the Right Owner | agnte](https://agnte.ai/platform/findings): Every finding from testers, reviews and agents is de-duplicated and routed to the role that should decide: fix now, later, not a bug, or needs clarification. - [Run Coding Agents Across Multiple Machines | agnte Fleet](https://agnte.ai/platform/fleet): Turn your PC and any SSH-reachable Linux VM into one fleet with a shared task pool, smart placement by real headroom, drain controls and safe CLI updates. - [Persistent AI Agent Memory That Never Compacts | agnte](https://agnte.ai/platform/memory): agnte stores decisions, lessons, knowledge and handoffs per project in a database shared by every machine, so agents never forget why code exists. - [AI Model Routing by Work Type | agnte Model Allocation](https://agnte.ai/platform/model-allocation): Map each work type to a model tier and effort, or pin it yourself. agnte flags expensive models on light work and weak models that keep failing. - [36 AI Agent Roles: A Full Engineering Team | agnte](https://agnte.ai/platform/roles): agnte runs 36 built-in roles, from business analyst and architect to QA, security, SRE and tech writer, with team presets and custom roles. - [Sandboxed AI Coding Agents with Central Policy | agnte](https://agnte.ai/platform/sandbox): Every agent runs sandboxed: bubblewrap and nftables on Linux, locked-down Docker on Windows, 73 denied commands, central policy for all machines and a full audit journal. - [Self-Healing AI Agent Operations | agnte](https://agnte.ai/platform/self-healing): A watchdog keeps work moving, outcome checks run every 15 minutes, and platform root causes become fix tasks with regression tests for your approval. - [Pool Multiple AI Coding Subscriptions into One | agnte](https://agnte.ai/platform/subscription-pool): Use several Claude Code subscriptions as one pool. Work moves to whichever account has quota, and a running task can continue on another account. - [AI Agent Budgets, Cost Forecast and Token Ledger | agnte](https://agnte.ai/platform/token-economy): Budgets with three warnings and a hard stop, a cost and ETA forecast per project, a per-request ledger and dashboards for every model, account and machine. - [Choose How Each Project Spends Tokens | agnte Token Strategies](https://agnte.ai/platform/token-strategies): Pick a token strategy per project: balanced across subscriptions, in priority order, subscriptions first with API as overflow, or the cheapest capable account. ## Surfaces - [Shared Knowledge Pool for AI Agents | agnte Agent Communication](https://agnte.ai/surfaces/agent-communication): agnte agents never message each other directly. They share questions, decisions and handoffs through one knowledge pool, and you can talk to any agent. - [Observe and Steer AI Coding Agents in Real Time | agnte Agent Control](https://agnte.ai/surfaces/agent-control): Watch every running agent across your fleet, read its live work, and give new instructions, pause or cancel it while it executes. Engineers stay in charge. - [agnte API, MCP Server and CLI: AI-Operable Agent Orchestration](https://agnte.ai/surfaces/api-mcp): Run agnte from code or from other agents: OpenAPI 3.1 with 110 scoped routes, an MCP server with 26 tools, llms.txt and a CLI with JSON output. - [Ask agnte: Built-in Assistant for Your AI Fleet](https://agnte.ai/surfaces/ask-agnte): Ask agnte explains alerts in plain English and changes settings for you after you confirm. Run your AI engineering team by asking. - [agnte Desktop App for Windows | Agent Orchestration Platform](https://agnte.ai/surfaces/desktop-app): Run the agnte control layer from a Windows desktop app with tray status, native notifications, a local store with automatic backups and an external guardian. - [GitHub Integration for AI Coding Agents | agnte](https://agnte.ai/surfaces/github): agnte uses GitHub as the source of truth: issues, Projects board, one pull request per task, role reviews, ADRs and two-way comments. Agents never hold your token. - [Needs-You Center: Human-in-the-Loop Approvals for AI Agents | agnte](https://agnte.ai/surfaces/needs-you): agnte calls you only when it matters: access, credentials, paid resources and protected approvals land in one Needs-You center. - [Live Steering: Steer, Pause and Cancel Running AI Agents | agnte](https://agnte.ai/surfaces/steering): Talk to a running coding agent, queue instructions, pause or cancel it. Live steering keeps you in charge of every agent session in your fleet. - [agnte Web Console: Kanban, Dashboards, Findings | Agent Orchestration](https://agnte.ai/surfaces/web-console): One browser workspace for your AI engineering team: Kanban with 8 lanes, live agent console, project map, cost dashboards, insights, findings and agent health. ## Automations - [Automated Accessibility Testing with axe-core | agnte](https://agnte.ai/automations/accessibility): agnte runs axe-core on every page its testers visit and routes accessibility findings to an accessibility role that decides each one. - [Automated QA for AI-Built Software: Test What Shipped | agnte](https://agnte.ai/automations/automated-qa): agnte deploys every build to a dev server with blue/green switching, runs API contract and log tests, and turns passing journeys into token-free tests. - [Bring AI Agents to an Existing Codebase | agnte Brownfield Import](https://agnte.ai/automations/brownfield): Import an existing codebase: agnte maps it, baselines its tests so old failures do not block, turns git history into lessons and triages open issues. - [Change Requests with Impact Analysis for AI Agents | agnte](https://agnte.ai/automations/change-requests): Ask for a feature, bug fix, refactor or test change. agnte shows the impact and exact task delta before you approve, and enforces a gate for each change type. - [AI Code Review by Role: Only the Right Reviewers | agnte](https://agnte.ai/automations/code-review): agnte ships one pull request per task and asks only the roles whose files changed to review it. An integrator role merges. Fewer reviews, better ones. - [AI Agent Incident Response and Root Cause | agnte](https://agnte.ai/automations/incident-response): agnte clusters agent failures by signature, model, host and role, raises incidents and investigates each one to a root cause with evidence. - [Reusable Agent Testing and Shared Test Memory | agnte Journey Replay](https://agnte.ai/automations/journey-replay): Agents record real user journeys once, save them as executable tests with expected outcomes, and replay them on every release. All agents share one test history. - [Living Documentation Maintained by AI Agents | agnte](https://agnte.ai/automations/living-docs): Tech writer and API docs roles own your documentation. Doc gaps become findings, so docs stay current with the code. - [AI Plan Pipeline: From Plain-English Idea to Approved Plan | agnte](https://agnte.ai/automations/plan-pipeline): Describe an idea in plain English. agnte runs it past analyst, product, project and architect roles, reviews the plan and waits for your approval. - [Automated Release Management for AI-Built Software | agnte](https://agnte.ai/automations/releases): agnte cuts milestone releases with a real git tag, a GitHub release and a changelog, after the product journeys gate passes. - [Scheduled Routines and Nightly AI Agent Runs | agnte](https://agnte.ai/automations/routines): Schedule heavy engineering work like security scans, soak tests and full test runs at night with human-readable routines, inside your budget. - [Automated Security Testing with OWASP ZAP | agnte](https://agnte.ai/automations/security-testing): agnte runs a passive OWASP ZAP baseline on your dev server every night, keeps a dependency audit baseline and routes every security finding to the security role.