llms.txt Content
# AgentCat
> Analytics for MCP servers. AgentCat tells you exactly what AI agents are doing on your product and where they're getting stuck.
AgentCat is an analytics platform for companies building MCP (Model Context Protocol) servers. It captures agent intent, replays sessions, tracks errors, and forwards telemetry to your existing observability tools.
## Key Features
- **Session Replay**: Step through exactly what agents did and why they made each tool call
- **Agent Intent Capture**: Understand *why* agents call your tools, not just what they call
- **Automatic Issue Detection**: Track errors, hallucinations, and problems agents encounter
- **Dashboard Analytics**: Visualize traffic by client, server, location, model, and more
- **Telemetry Forwarding**: Send enriched data to Datadog, Sentry, PostHog via OpenTelemetry
- **Privacy Controls**: Client-side and server-side redaction, encryption at rest
## SDKs
Available for TypeScript/JavaScript, Python, and Go:
- npm: `npm i agentcat`
- pip: `pip install agentcat`
- go: `go get github.com/mcpcat/mcpcat-go-sdk/officialsdk@latest`
## Quick Start
```typescript
import * as agentcat from "agentcat";
agentcat.track(mcpServer, "proj_YOUR_PROJECT_ID");
```
## Documentation
- [Full Documentation](https://docs.agentcat.com)
- [LLMs.txt (docs index)](https://docs.agentcat.com/llms.txt)
- [LLMs-full.txt (complete docs)](https://docs.agentcat.com/llms-full.txt)
## Doc Pages
- [Introduction](https://docs.agentcat.com/): Overview of AgentCat and what it helps with
- [Quickstart](https://docs.agentcat.com/quickstart): Install the AgentCat SDK in 1 minute
- [How It Works](https://docs.agentcat.com/how-it-works): Architecture, agent intent capture, request tracing, sessions
- [Identifying Users](https://docs.agentcat.com/sdk/identifying-users): Associate sessions with your users
- [Tool Call Context](https://docs.agentcat.com/sdk/tool-call-context): Customize what context AgentCat captures
- [Debug Mode](https://docs.age