# Intertrace Documentation

> Protect AI with AI. Inline security gateway + posture dashboard.

## Docs

### Introduction

- [Overview](https://docs.intertrace.ai/docs/overview): What Intertrace is and is not
- [How Intertrace works](https://docs.intertrace.ai/docs/how-it-works): Data plane + control plane
- [FAQ](https://docs.intertrace.ai/docs/faq)

### Getting started

- [Quickstart](https://docs.intertrace.ai/docs/quickstart): Go live in minutes
- [Set up with AI](https://docs.intertrace.ai/docs/ai-setup): One prompt for your coding agent
- [Early access & signup](https://docs.intertrace.ai/docs/early-access): Waitlist, access codes, first org
- [Org invites](https://docs.intertrace.ai/docs/org-invites): Invite teammates to your workspace
- [Authentication](https://docs.intertrace.ai/docs/auth): Runtime keys vs org API keys
- [Control plane](https://docs.intertrace.ai/docs/control-plane): itr CLI, coding-agent hooks, policy gitops, approvals
- [Connection methods](https://docs.intertrace.ai/docs/connection-methods): How traffic reaches Intertrace

### Connect clients

- [OpenAI SDK](https://docs.intertrace.ai/docs/connect/openai-sdk)
- [Coding agents](https://docs.intertrace.ai/docs/coding-agents): What itr connect governs on each host
- [Claude Code](https://docs.intertrace.ai/docs/connect/claude-code)
- [Cursor](https://docs.intertrace.ai/docs/connect/cursor)
- [Claude Desktop](https://docs.intertrace.ai/docs/connect/claude-desktop)
- [Devin](https://docs.intertrace.ai/docs/connect/devin)
- [Codex](https://docs.intertrace.ai/docs/connect/codex)
- [OpenCode](https://docs.intertrace.ai/docs/connect/opencode)
- [GitHub Copilot](https://docs.intertrace.ai/docs/connect/github-copilot)
- [Microsoft Copilot Studio](https://docs.intertrace.ai/docs/connect/copilot-studio): Beta: discovery and inline protection
- [Claude Enterprise](https://docs.intertrace.ai/docs/connect/claude-enterprise): Beta: inference hooks for every Claude app
- [LiteLLM](https://docs.intertrace.ai/docs/connect/litellm)
- [Open WebUI](https://docs.intertrace.ai/docs/connect/open-webui)
- [MCP servers](https://docs.intertrace.ai/docs/connect/mcp)
- [Product MCP](https://docs.intertrace.ai/docs/mcp)
- [Agent frameworks](https://docs.intertrace.ai/docs/connect/frameworks)
- [ServiceNow AI agents](https://docs.intertrace.ai/docs/connect/servicenow): Beta: discover agents, guard their tools
- [Salesforce Agentforce](https://docs.intertrace.ai/docs/connect/salesforce-agentforce): Beta: discover agents, guard actions
- [GitLab](https://docs.intertrace.ai/docs/connect/gitlab): Beta: projects, CI agents and keys in Discovery
- [Browser extension](https://docs.intertrace.ai/docs/connect/browser-extension): Shadow AI: visits, and checks before people send to AI apps
- [Compliance APIs](https://docs.intertrace.ai/docs/connect/compliance-apis): Beta: Claude, ChatGPT, Microsoft 365 Copilot and n8n in Discovery
- [curl / raw HTTP](https://docs.intertrace.ai/docs/connect/curl)
- [Models & providers](https://docs.intertrace.ai/docs/connect/models)
- [A2A, service & inference](https://docs.intertrace.ai/docs/planes): Every plane the gateway governs

### Governance

- [The agent page](https://docs.intertrace.ai/docs/agent-page): Connections, data, approvals, limits, activity
- [Call limits](https://docs.intertrace.ai/docs/call-limits): At most N calls to a tool or server per conversation, hour or day
- [Roles & consequence tiers](https://docs.intertrace.ai/docs/roles-tiers)
- [Learning period](https://docs.intertrace.ai/docs/learning-period): Learn, review, then lock in
- [Approvals](https://docs.intertrace.ai/docs/approvals): Inbox, channels, plans, scopes
- [Confirm in client](https://docs.intertrace.ai/docs/confirm-in-client): Low-friction asks for developers
- [Guardrails & proposals](https://docs.intertrace.ai/docs/guardrails)
- [Tool profiles](https://docs.intertrace.ai/docs/tool-profiles): How the gateway reads what a call does
- [Policy packs & industries](https://docs.intertrace.ai/docs/policy-packs)

### Visibility

- [Discovery](https://docs.intertrace.ai/docs/discovery): Find the AI already in use
- [Sessions & traces](https://docs.intertrace.ai/docs/sessions-traces)
- [Activity ledger](https://docs.intertrace.ai/docs/activity-ledger): What agents actually touched
- [MCP threat intel](https://docs.intertrace.ai/docs/mcp-threat-intel)
- [HIRO security analyst](https://docs.intertrace.ai/docs/hiro)

### Dashboard

- [Product spine](https://docs.intertrace.ai/docs/dashboard/product-spine): Connect → Estate → … → Settings
- [Capabilities map](https://docs.intertrace.ai/docs/dashboard-capabilities)
- [Estate](https://docs.intertrace.ai/docs/dashboard/estate)
- [Traffic](https://docs.intertrace.ai/docs/dashboard/traffic)
- [Events](https://docs.intertrace.ai/docs/dashboard/events)
- [Work](https://docs.intertrace.ai/docs/dashboard/work)
- [Guardian](https://docs.intertrace.ai/docs/guardian-production): Watches your estate, recommends next actions
- [Graph](https://docs.intertrace.ai/docs/dashboard/graph)
- [MCP console](https://docs.intertrace.ai/docs/dashboard/mcp)
- [Policies](https://docs.intertrace.ai/docs/dashboard/policies)
- [Labs (Red Team)](https://docs.intertrace.ai/docs/dashboard/labs)
- [Labs in CI](https://docs.intertrace.ai/docs/labs/ci): Red-team an agent on every pull request; SARIF to code scanning
- [Costs](https://docs.intertrace.ai/docs/dashboard/costs)
- [Response cache](https://docs.intertrace.ai/docs/response-cache): Answer repeated requests without a provider call
- [Your own kinds of data](https://docs.intertrace.ai/docs/data-classes): Employee IDs, codenames, account numbers
- [Outside guardrails](https://docs.intertrace.ai/docs/external-guardrails): Your own screening service or Lakera Guard as a second opinion
- [Grounding check](https://docs.intertrace.ai/docs/grounding): Flag or withhold replies their sources don't support
- [Settings](https://docs.intertrace.ai/docs/dashboard/settings)

### Guides

- [Secure your MCP](https://docs.intertrace.ai/docs/mcp/secure-your-mcp): Put Intertrace in front of MCP servers
- [Policy setup](https://docs.intertrace.ai/docs/policy)
- [Behavioral monitoring](https://docs.intertrace.ai/docs/behavioral-monitoring)
- [Intent alignment](https://docs.intertrace.ai/docs/intent-alignment)
- [Findings & incidents](https://docs.intertrace.ai/docs/detections-incidents)
- [Webhook integration](https://docs.intertrace.ai/docs/webhooks)
- [AWS discovery IAM](https://docs.intertrace.ai/docs/cloud-discovery-iam-policy)
- [Troubleshooting](https://docs.intertrace.ai/docs/troubleshooting)

### Operations & account

- [Integrations](https://docs.intertrace.ai/docs/integrations): What each one sends, how to test it, what it doesn't do yet
- [SIEM & audit exports](https://docs.intertrace.ai/docs/siem-exports)
- [Sign-in, SCIM & API keys](https://docs.intertrace.ai/docs/sso-scim)
- [Agent identities](https://docs.intertrace.ai/docs/delegated-authority): Beta
- [Plans & usage](https://docs.intertrace.ai/docs/plans-usage)
- [Status & health](https://docs.intertrace.ai/docs/status-health)
- [Customer-hosted gateway](https://docs.intertrace.ai/docs/customer-hosted): Pilot

### API & SDKs

- [API reference](https://docs.intertrace.ai/docs/api)
- [Event schema](https://docs.intertrace.ai/docs/events/schema)
- [Quickstart: Python](https://docs.intertrace.ai/docs/quickstart/python)
- [Quickstart: Node](https://docs.intertrace.ai/docs/quickstart/node)
- [SDKs](https://docs.intertrace.ai/docs/sdks): Public npm runtime + tracing

## Set up Intertrace with a coding agent

- [Set up with AI](https://docs.intertrace.ai/docs/ai-setup): the page with the one-click copy.
- [agent-setup.txt](https://docs.intertrace.ai/agent-setup.txt): step-by-step instructions for a coding agent (Claude Code, Cursor, Codex, Copilot) to put a project's AI agent behind Intertrace, and the block to keep in the repository's AGENTS.md so later edits keep it there. The connect screen (Connect → An agent you built) gives the same prompt with your agent's values filled in.

## For agents

An agent asks Intertrace before it acts. These run on the gateway (https://intertrace.fly.dev) with the agent's own runtime key (Bearer itr_rt_...):

- POST /v1/actions/authorize: before a tool call, send the action; the answer is ALLOW, BLOCK or REQUIRE_APPROVAL with reason codes.
- GET /v1/runtime/approvals/{id}?wait=25: after REQUIRE_APPROVAL, wait for a person's decision; once approved, call authorize again with approval_id and the same action.
- POST /v1/runtime/plans: one approval for the steps of a task in a session.
- /mcp: the product MCP server (OAuth discovery at /.well-known/oauth-protected-resource/mcp).

## OpenAPI

- [openapi.json](https://docs.intertrace.ai/openapi.json)