# Agent Sentinel

## Docs

- [Actions & runs](https://docs.agentsentinel.dev/concepts/actions-and-runs.md): How Agent Sentinel models agent execution: runs contain actions.
- [Ledger](https://docs.agentsentinel.dev/concepts/ledger.md): Local-first JSONL log format written by the SDK.
- [Policies](https://docs.agentsentinel.dev/concepts/policies.md): Budgets, allow/deny lists, rate limits, and remote policy sync.
- [Agents Page](https://docs.agentsentinel.dev/console/agents.md): Discover and manage your agent fleet with automatic registration and performance metrics.
- [Analytics Page](https://docs.agentsentinel.dev/console/analytics.md): Cost and performance analytics with visual dashboards and trend analysis.
- [Approvals Page](https://docs.agentsentinel.dev/console/approvals.md): Manage human-in-the-loop approval workflows for sensitive agent actions.
- [Determinism Analysis](https://docs.agentsentinel.dev/console/determinism-analysis.md): Analyze runs for non-deterministic patterns and get actionable recommendations.
- [Evals](https://docs.agentsentinel.dev/console/evals.md): Benchmark scorecards, regression tracking, CI gating, and adversarial scenario generation.
- [Interventions Page](https://docs.agentsentinel.dev/console/interventions.md): Monitor policy enforcement and track what Agent Sentinel blocks in real-time.
- [Kill Switch](https://docs.agentsentinel.dev/console/kill-switch.md): Top-level emergency controls for halting agents, runs, or missions.
- [MCP Explorer](https://docs.agentsentinel.dev/console/mcp.md): Browse the platform's Model Context Protocol surface — tools, resources, and prompts.
- [Web Console Overview](https://docs.agentsentinel.dev/console/overview.md): Monitor, manage, and control your AI agents through the Agent Sentinel web console.
- [Console Quick Start](https://docs.agentsentinel.dev/console/quickstart.md): Get started with the Agent Sentinel web console in 5 minutes.
- [Replay Hub](https://docs.agentsentinel.dev/console/replay.md): Org-wide history of local-replay submissions from every SDK client.
- [Runs Page](https://docs.agentsentinel.dev/console/runs.md): View execution history, replay runs, and analyze determinism.
- [Settings Page](https://docs.agentsentinel.dev/console/settings.md): Configure policies, manage API keys, and set up notifications.
- [Development](https://docs.agentsentinel.dev/development.md): Run the platform, web console, and docs locally.
- [Agent Sentinel](https://docs.agentsentinel.dev/index.md): Kill switch, observability, and cost control for Python AI agents.
- [Agent Discovery](https://docs.agentsentinel.dev/platform/agents.md): Automatic agent registration and metrics tracking in the platform.
- [Approvals API](https://docs.agentsentinel.dev/platform/approvals.md): Human-in-the-loop approval workflows for sensitive agent actions.
- [Authentication](https://docs.agentsentinel.dev/platform/auth.md): API keys for SDK integration and JWT tokens for web console access.
- [Compliance (Enterprise foundations)](https://docs.agentsentinel.dev/platform/compliance.md): EU AI Act export and oversight endpoints (foundation).
- [Evals API](https://docs.agentsentinel.dev/platform/evals.md): Submit benchmark results, compare runs for regressions, and generate adversarial scenarios.
- [Interventions API](https://docs.agentsentinel.dev/platform/interventions.md): Track when Agent Sentinel blocks, modifies, or escalates actions to demonstrate platform value.
- [Kill Switch API](https://docs.agentsentinel.dev/platform/kill-switch.md): Emergency endpoints to halt agents, runs, or missions, and check kill state from the SDK.
- [Activity Ledger API](https://docs.agentsentinel.dev/platform/ledger.md): Query the immutable audit trail of all agent actions
- [MCP (Model Context Protocol)](https://docs.agentsentinel.dev/platform/mcp.md): LLM-native API for autonomous platform management via structured tool calls.
- [Platform API Overview](https://docs.agentsentinel.dev/platform/overview.md): The Agent Sentinel cloud platform API for ingesting telemetry and querying data.
- [Policies API](https://docs.agentsentinel.dev/platform/policies.md): Create policies, compile from prose, manage version history, rollback, and sync to SDKs.
- [Replay (platform)](https://docs.agentsentinel.dev/platform/replay.md): Replay analysis endpoints for determinism and divergence detection.
- [Runs & actions](https://docs.agentsentinel.dev/platform/runs.md): Query runs and their actions from the platform.
- [Stats](https://docs.agentsentinel.dev/platform/stats.md): Aggregate cost and reliability metrics for dashboards.
- [WebSocket Real-time Updates](https://docs.agentsentinel.dev/platform/websockets.md): Real-time event streaming for live dashboard updates and notifications.
- [Architecture](https://docs.agentsentinel.dev/project/architecture.md): The canonical architecture overview for Agent Sentinel.
- [Deployment](https://docs.agentsentinel.dev/project/deployment.md)
- [Launch Path](https://docs.agentsentinel.dev/project/launch-path.md): The focused launch plan for Agent Sentinel.
- [Roadmap](https://docs.agentsentinel.dev/project/roadmap.md): Future product direction without scattering plan documents.
- [Quickstart](https://docs.agentsentinel.dev/quickstart.md): Instrument your agent in minutes, with local-first logs and optional platform sync.
- [Approval Workflows](https://docs.agentsentinel.dev/sdk/python/approvals.md): Implement human-in-the-loop approval workflows for sensitive agent actions.
- [Compliance (Enterprise foundations)](https://docs.agentsentinel.dev/sdk/python/compliance.md): Human-in-the-loop metadata and EU AI Act export endpoints (foundation).
- [Error Handling](https://docs.agentsentinel.dev/sdk/python/errors.md): Understand Agent Sentinel's error types and how to handle them gracefully.
- [Framework Integrations](https://docs.agentsentinel.dev/sdk/python/framework-integrations.md): Integrate Agent Sentinel with LangChain, CrewAI, AutoGen, and other agent frameworks.
- [Guardrails](https://docs.agentsentinel.dev/sdk/python/guardrails.md): First-class runtime primitives: PII detection, content moderation, idempotency, and loop protection.
- [Installation](https://docs.agentsentinel.dev/sdk/python/installation.md): Install the Agent Sentinel Python SDK.
- [Instrumentation](https://docs.agentsentinel.dev/sdk/python/instrumentation.md): Wrap tool calls with telemetry and cost tracking using @guarded_action.
- [Interventions](https://docs.agentsentinel.dev/sdk/python/interventions.md): Track when Agent Sentinel blocks, modifies, or escalates actions to demonstrate platform value.
- [LLM Integrations](https://docs.agentsentinel.dev/sdk/python/llm-integrations.md): Automatically track LLM API calls with transparent instrumentation for OpenAI, Anthropic, Grok, and Gemini.
- [MCP Integration](https://docs.agentsentinel.dev/sdk/python/mcp.md): Use the Model Context Protocol to give LLMs direct access to Agent Sentinel platform features.
- [Policies & budgets](https://docs.agentsentinel.dev/sdk/python/policies.md): Enforce cost limits, allow/deny lists, and rate limits in the SDK.
- [Prose policies](https://docs.agentsentinel.dev/sdk/python/prose-policies.md): Compile YAML, JSON, or plain-English policy descriptions into the platform's policy IR.
- [Remote sync](https://docs.agentsentinel.dev/sdk/python/remote-sync.md): Upload local ledger entries to the Agent Sentinel Platform in the background.
- [Replay](https://docs.agentsentinel.dev/sdk/python/replay.md): Mock tool calls with recorded outputs to debug and reduce cost.

## OpenAPI Specs

- [openapi](https://docs.agentsentinel.dev/archive/mintlify-starter/api-reference/openapi.json)

## Optional

- [GitHub](https://github.com/agent-sentinel/)
- [PyPI](https://pypi.org/project/agentsentinel-sdk/)