Open Source
Tools for the AI agent stack.
Free, MIT-licensed products that make AI agents safe, accountable, and benchmarkable in real enterprises. Run them yourself — or talk to us about deploying them at scale.
Observer
Observability for AI coding agents. See what Claude Code, Codex, and Cursor are doing and what they're costing — without sending a byte off your machine. Multi-agent. Privacy-first. Self-hostable.
TypeScriptMIT
View product
db-mcp
A safe way to query company data with Claude. An MCP server with a self-improving semantic layer that gets smarter with every query. Postgres, ClickHouse, Trino, MySQL, SQL Server.
PythonMIT
View product
mcp-gateway-bench
The reference benchmark for MCP gateways. Auth, policy, audit, and latency — replayed through every gateway against a shared upstream. Publicly reproducible.
TypeScriptOpen source
View product