# AutoPIL

## Documentation

- [AutoPIL Documentation](https://autopil-docs.documentationai.com/Home.md): Runtime governance for autonomous AI agents — enforce data-access policy at the retrieval layer and prove every decision with a tamper-evident audit trail.
- [Quick start](https://autopil-docs.documentationai.com/getting-started/quickstart.md): Install the AutoPIL Python package, point it at a policy directory, and protect your first retrieval function.
- [Installation](https://autopil-docs.documentationai.com/getting-started/installation.md): Install options for the SQLite, Postgres, and OpenTelemetry backends, with a first protected-retrieval example.
- [First-start bootstrap](https://autopil-docs.documentationai.com/getting-started/first-start.md): Start the API server, capture the one-time superadmin key, and recover it if you miss it.
- [MCP Server](https://autopil-docs.documentationai.com/getting-started/mcp.md): Connect AI assistants to the AutoPIL documentation using the Model Context Protocol (MCP).
- [Architecture](https://autopil-docs.documentationai.com/core-concepts/architecture.md): Where AutoPIL sits in the stack — governing retrieval between your data sources and agents.
- [Enforcement flow](https://autopil-docs.documentationai.com/core-concepts/enforcement-flow.md): The ten ordered checks every guard.protect() call runs, from agent identity binding to policy evaluation.
- [Multi-tenancy](https://autopil-docs.documentationai.com/core-concepts/multi-tenancy.md): How row-level tenant isolation scopes audit events, policies, and keys to each tenant.
- [Agent identity](https://autopil-docs.documentationai.com/core-concepts/agent-identity.md): Layered agent identity enforcement — registry approval, role restriction, key binding, and policy-level agent IDs.
- [Policy file format](https://autopil-docs.documentationai.com/policy-guide/policy-format.md): Structure of an AutoPIL policy YAML file and how multiple roles are defined.
- [Fields reference](https://autopil-docs.documentationai.com/policy-guide/fields-reference.md): Reference for every policy field — its type, whether it is required, and what it controls.
- [Evaluation order](https://autopil-docs.documentationai.com/policy-guide/evaluation-order.md): The order policy conditions are checked and where evaluation stops on the first match.
- [Sensitivity levels](https://autopil-docs.documentationai.com/policy-guide/sensitivity-levels.md): The four sensitivity levels from low to critical and the data each is intended for.
- [Task scoping](https://autopil-docs.documentationai.com/policy-guide/task-scoping.md): Restrict a role to categories of work using task_type with allowed_tasks and denied_tasks.
- [Hot reload](https://autopil-docs.documentationai.com/policy-guide/hot-reload.md): Apply edited policy files live — no server restart and no interruption to running agents.
- [Agent ID binding in policies](https://autopil-docs.documentationai.com/policy-guide/agent-id-binding.md): Lock a policy to specific registered agent instances with permitted_agent_ids.
- [Choosing an integration pattern](https://autopil-docs.documentationai.com/sdks/integration-patterns.md): Choose between the embedded decorator and the REST API for your deployment and identity model.
- [Python SDK](https://autopil-docs.documentationai.com/sdks/python-sdk.md): Embed the ContextGuard decorator directly in your Python agent process.
- [Connecting to a hosted instance (Render / production)](https://autopil-docs.documentationai.com/sdks/connecting-to-hosted-instance.md): Call the REST API from Python when AutoPIL runs on Render or another remote host.
- [Async decorator — protect_async](https://autopil-docs.documentationai.com/sdks/async-protect-async.md): Use protect_async for FastAPI, asyncio, and other async frameworks with contextvars support.
- [ASGI Middleware](https://autopil-docs.documentationai.com/sdks/middleware.md): Enforce access policy at the HTTP layer for FastAPI and Starlette with AutoPILMiddleware.
- [MCP Server](https://autopil-docs.documentationai.com/sdks/mcp-server.md): Expose AutoPIL context governance as native MCP tools for Claude, GPT, Gemini, and other agents.
- [Framework integrations](https://autopil-docs.documentationai.com/sdks/frameworks.md): Native guard subclasses for the five major agent frameworks, each tagged with its own source_type.
- [Go SDK](https://autopil-docs.documentationai.com/sdks/go-sdk.md): A dependency-free Go REST client built on net/http and encoding/json, requiring Go 1.21+.
- [Java SDK](https://autopil-docs.documentationai.com/sdks/java-sdk.md): A Jackson-based Java REST client on java.net.http.HttpClient, requiring Java 17+.
- [Local development](https://autopil-docs.documentationai.com/deployment/local.md): Run the core backend and dashboard from the monorepo in local development, with a server flag reference.
- [Docker](https://autopil-docs.documentationai.com/deployment/docker.md): Run AutoPIL with Docker, including the admin-key lookup.
- [Postgres](https://autopil-docs.documentationai.com/deployment/postgres.md): Point AutoPIL at Postgres, with automatic schema creation and safe idempotent migrations.
- [OpenTelemetry](https://autopil-docs.documentationai.com/deployment/opentelemetry.md): Enable OpenTelemetry traces and metrics; a complete no-op when the OTLP endpoint is unset.
- [Alerting (SMTP)](https://autopil-docs.documentationai.com/deployment/alerting.md): Configure SMTP email delivery for alert rules and how missing settings are handled.
- [Retention](https://autopil-docs.documentationai.com/deployment/retention.md): Set per-tenant audit and session retention windows from the dashboard in production.
- [PII masking](https://autopil-docs.documentationai.com/deployment/pii-masking.md): Mask sensitive fields in audit events before they are written, for privacy-sensitive deployments.

## API Reference

- [Authentication](https://autopil-docs.documentationai.com/api-reference/authentication.md): API key authentication for all /v1/ routes via the X-API-Key header, including 401 and 403 behavior.
- [Session retention](https://autopil-docs.documentationai.com/api-reference/session-retention.md): How expired and revoked session records are purged independently of audit events.
- [PII masking](https://autopil-docs.documentationai.com/api-reference/pii-masking.md): Which audit fields are masked when PII masking is enabled, and how existing events are treated.
- [Evaluate a retrieval request](https://autopil-docs.documentationai.com/api-reference/context/evaluate-context.md): Evaluate a retrieval request against policy. Returns ALLOW or DENY with the matched policy name, a human-readable reason, and an event_id for lineage.
- [List audit events](https://autopil-docs.documentationai.com/api-reference/audit/list-audit-events.md): List audit events, filterable by role, decision, and source.
- [Full audit trail for a session](https://autopil-docs.documentationai.com/api-reference/audit/get-session-audit.md): Return all events for a session, in order.
- [Aggregate audit stats](https://autopil-docs.documentationai.com/api-reference/audit/get-audit-stats.md): Aggregate stats — totals, deny rate, top roles, and a by_source_type breakdown distinguishing SDK, API, framework, and REST traffic.
- [Verify hash-chain integrity](https://autopil-docs.documentationai.com/api-reference/audit/verify-audit-chain.md): Verify the SHA-256 hash chain integrity for this tenant.
- [Retention policy and last purge anchor](https://autopil-docs.documentationai.com/api-reference/audit/get-audit-retention.md): Active retention policy and last purge anchor for this tenant. anchor is null if no purge has run yet.
- [Record a downstream action](https://autopil-docs.documentationai.com/api-reference/audit/record-lineage.md): Record a downstream action linked to an audit event (lineage).
- [Event and its downstream actions](https://autopil-docs.documentationai.com/api-reference/audit/get-lineage.md): Return the event and all linked downstream actions.
- [Log retention](https://autopil-docs.documentationai.com/api-reference/settings/log-retention.md): Configure per-tenant audit and session retention windows and inspect the purge chain anchor.
- [Get retention settings](https://autopil-docs.documentationai.com/api-reference/settings/get-retention-settings.md): Return this tenant's current audit and session retention settings.
- [Update retention settings](https://autopil-docs.documentationai.com/api-reference/settings/update-retention-settings.md): Update retention settings (admin-scoped key required). Either field is optional — omit one to leave it unchanged. Set to 0 to disable automatic purge for that data type.
- [List active policies](https://autopil-docs.documentationai.com/api-reference/policies/list-policies.md): List active policies, filterable by industry, process group, or role.
- [Create a policy](https://autopil-docs.documentationai.com/api-reference/policies/create-policy.md): Create a policy
- [Update a policy](https://autopil-docs.documentationai.com/api-reference/policies/update-policy.md): Update a policy (saves version history, sets user_modified).
- [Soft-delete a policy](https://autopil-docs.documentationai.com/api-reference/policies/delete-policy.md): Soft-delete a policy
- [Policy version history](https://autopil-docs.documentationai.com/api-reference/policies/get-policy-history.md): Full version history for a policy.
- [Restore a policy](https://autopil-docs.documentationai.com/api-reference/policies/restore-policy.md): Restore a policy to its original YAML definition (clears user_modified).
- [Reload policies from disk](https://autopil-docs.documentationai.com/api-reference/policies/reload-policies.md): Reload all policies from disk without restart.
- [Alert rule types](https://autopil-docs.documentationai.com/api-reference/alerts/alert-rule-types.md): Define threshold alert rules and review the deliveries they fire.
- [Create an alert rule](https://autopil-docs.documentationai.com/api-reference/alerts/create-alert-rule.md): Create an alert rule. Rule types: denial_spike, new_source_access, isolation_violation, high_deny_rate.
- [List alert rules](https://autopil-docs.documentationai.com/api-reference/alerts/list-alert-rules.md): List alert rules
- [Update an alert rule](https://autopil-docs.documentationai.com/api-reference/alerts/update-alert-rule.md): Update a rule (enable/disable, thresholds).
- [Delete an alert rule](https://autopil-docs.documentationai.com/api-reference/alerts/delete-alert-rule.md): Delete an alert rule
- [List fired alert deliveries](https://autopil-docs.documentationai.com/api-reference/alerts/list-fired-alerts.md): List fired alert deliveries with trigger details.
- [List sessions](https://autopil-docs.documentationai.com/api-reference/sessions/list-sessions.md): List sessions, filterable by role and status.
- [Session detail](https://autopil-docs.documentationai.com/api-reference/sessions/get-session.md): Session detail and status.
- [Revoke a session](https://autopil-docs.documentationai.com/api-reference/sessions/revoke-session.md): Revoke a session immediately.
- [API key scopes](https://autopil-docs.documentationai.com/api-reference/keys/key-scopes.md): Create, list, rotate, revoke, and bind API keys across three permission scopes.
- [Key binding](https://autopil-docs.documentationai.com/api-reference/keys/key-binding.md): Bind a key to a registered agent and restrict it to specific roles.
- [Create an API key](https://autopil-docs.documentationai.com/api-reference/keys/create-key.md): Create an API key — the plaintext key is returned once; store it immediately. Keys support three scopes: admin, read, evaluate.
- [List keys](https://autopil-docs.documentationai.com/api-reference/keys/list-keys.md): List keys for the tenant (no plaintext).
- [Revoke a key](https://autopil-docs.documentationai.com/api-reference/keys/revoke-key.md): Revoke a key immediately.
- [Rotate a key](https://autopil-docs.documentationai.com/api-reference/keys/rotate-key.md): Rotate a key — creates a replacement with the same scope and expiry, and revokes the original.
- [Bind a key to an agent or roles](https://autopil-docs.documentationai.com/api-reference/keys/bind-key.md): Bind a key to a registered agent and/or restrict it to specific roles. Both fields are optional and independent.
- [Agent registry](https://autopil-docs.documentationai.com/api-reference/agents/registry-overview.md): Register, approve, and manage agent instances in the authoritative agent registry.
- [Register a new agent](https://autopil-docs.documentationai.com/api-reference/agents/register-agent.md): Register a new agent — status starts as draft.
- [List agents](https://autopil-docs.documentationai.com/api-reference/agents/list-agents.md): List all registered agents for this tenant.
- [Get an agent](https://autopil-docs.documentationai.com/api-reference/agents/get-agent.md): Get a single agent by ID — includes live activity stats.
- [Update agent metadata](https://autopil-docs.documentationai.com/api-reference/agents/update-agent.md): Update agent metadata — display name, description, framework, owner.
- [Transition agent approval status](https://autopil-docs.documentationai.com/api-reference/agents/update-agent-status.md): Transition approval status: draft → pending_approval → approved → deprecated.
- [Registry-wide agent counts](https://autopil-docs.documentationai.com/api-reference/agents/get-agent-stats-summary.md): Registry-wide counts — total, approved, pending, deny rates.
- [Tenant enforcement settings](https://autopil-docs.documentationai.com/api-reference/admin/tenant-enforcement.md): Superadmin-only per-tenant enforcement flags, such as requiring an agent_id.
- [Superadmin key provisioning](https://autopil-docs.documentationai.com/api-reference/admin/superadmin-provisioning.md): Provision keys for any tenant as a superadmin, without that tenant's credentials.
- [Create a tenant](https://autopil-docs.documentationai.com/api-reference/admin/create-tenant.md): Create a tenant. Superadmin key only.
- [Update tenant enforcement settings](https://autopil-docs.documentationai.com/api-reference/admin/update-tenant-settings.md): Update per-tenant enforcement flags — superadmin only.
- [Provision a key for any tenant](https://autopil-docs.documentationai.com/api-reference/admin/provision-tenant-key.md): Provision a key for any tenant — superadmin only. Accepts the same name, scope, and expires_days fields as POST /v1/keys.
- [Liveness probe](https://autopil-docs.documentationai.com/api-reference/health/health.md): Health check. No authentication required.