Latitude
An open-source AI agent monitoring platform under MIT. Self-host with Docker or Kubernetes, track failures, and dispatch Claude Code or Cursor to fix them.
Latitude is a self-hosted AI agent monitoring platform for teams that need to find production failures and check that fixes work. It connects recurring problems to the sessions that caused them, then can dispatch Claude Code or Cursor with enough context to make a fix and open a pull request.
You can run the MIT-licensed software on your own server with Docker Compose or Kubernetes. A hosted service is also available; it stores and processes hosted data in Europe. Self-hosting lets you keep the monitoring infrastructure under your control, while your application's model provider remains a separate choice.
Latitude captures multi-turn conversations, tool calls, errors, costs, and execution paths. It accepts OpenTelemetry traces and works with OpenAI, Anthropic, Bedrock, Vercel AI SDK, and LangChain. Its conversation analysis identifies events such as abandoned sessions, retries, escalations, and tool failures.
Similar failures become tracked issues with examples and trends, so teams can investigate recurring behavior rather than read individual logs. Custom signals track specific failure modes, policies, or quality checks. Alerts reach Slack, email, or webhooks when an issue appears or escalates.
Semantic search and metadata filters help narrow down affected sessions. Teams can annotate traces and outputs, then use validated production traces to build versioned datasets for evaluations and regression tests. Latitude replays fixes against failing traces, and its MCP server and CLI give coding agents access to the workspace's monitoring and project controls.
