# Agent Basin

> Observability and replay platform for AI agent sessions.

Agent Basin captures every tool call, decision branch, and output from AI agent runs in production. Runs are stored, searchable, and fully replayable — giving engineering teams a complete flight recorder for their agents.

## Product

- **Every agent run captured**: Tool calls, LLM prompts, branch decisions, outputs — all stored with microsecond timestamps.
- **Step-by-step replay**: Replay any run exactly as it happened. Pause, rewind, inspect state at each decision point.
- **Searchable history**: Query runs by tool, error, duration, cost, or any custom attribute.
- **Failure alerts**: Instant notifications via Slack or PagerDuty when runs fail or hit new error patterns.
- **Branch diffing**: Compare two runs side-by-side to see where their decision trees diverged.
- **Cost & token tracking**: Per-run token usage and cost breakdown by step, tool, and model.

## Supported Frameworks

LangChain, AutoGPT, CrewAI, LlamaIndex, OpenAI Assistants, custom Python frameworks.

## Pricing

- **Free**: 1M events/month, 7-day retention, 1 project
- **Growth**: $99/month, 10M events/month, 90-day retention, unlimited projects, failure alerts, branch diffing
- **Enterprise**: Custom pricing, unlimited events, custom retention, SSO, self-hosted option

## Key Pages

- [Home](https://agentbasin.com/)
- [Pricing](https://agentbasin.com/pricing.html)
- [Sign Up / Early Access](https://agentbasin.com/signup.html)
- [About](https://agentbasin.com/about.html)
- [Blog](https://agentbasin.com/blog/)
- [Privacy Policy](https://agentbasin.com/privacy.html)
- [Terms of Service](https://agentbasin.com/terms.html)

## Blog Articles

- [Why your AI agent failed last Tuesday (and why you'll never know)](https://agentbasin.com/blog/why-your-ai-agent-failed.html)
- [The hidden cost of unobservable agents in production](https://agentbasin.com/blog/hidden-cost-unobservable-agents.html)
- [LangChain vs AutoGPT: which frameworks give you the best logs](https://agentbasin.com/blog/langchain-vs-autogpt-logs.html)
- [How to build a post-mortem process for AI agent incidents](https://agentbasin.com/blog/ai-agent-post-mortem-process.html)
- [Agent observability vs model observability: what most teams get wrong](https://agentbasin.com/blog/agent-vs-model-observability.html)

## Contact

- General: hello@agentbasin.com
- Press: press@agentbasin.com
- Privacy: privacy@agentbasin.com
- Careers: careers@agentbasin.com