obot

Why we included this project

Obot is for teams that need to give developers and their AI agents safe, controlled access to models and MCP tools without scattering provider credentials across the org. It sits between clients and the outside world as a governed layer: an MCP gateway proxies and filters tool access, and an LLM gateway exposes provider-compatible endpoints so approved models can be reached through services like OpenAI, Anthropic, Amazon Bedrock, and Azure. Around that it handles identity and permissions, stores secrets centrally, runs hosted MCP servers in sandboxes, and keeps correlated audit logs of AI activity across servers and end-user devices. It is intentionally not tied to one AI client or provider, so desktop tools like Claude Code, Codex, Cursor, and VS Code can adopt just the parts they need. The usual pain point it removes is sprawl: unmanaged MCP servers, credentials kept in different places, and little visibility into what agents actually do.