Centaur is an open-source platform from Paradigm that gives teams a shared AI agent they can talk to in Slack and trust with real work. Hosted on GitHub with about 1,400 stars and under very active development through 2026, it describes itself as frontier, agentic infrastructure that you own — an open-source alternative to hosted team agents.
Instead of every teammate running a separate local assistant, Centaur runs one shared agent per team. You mention the bot in a Slack thread, and the agent works inside an isolated Kubernetes sandbox equipped with a shell, a workspace, git, Python, Node.js and common development tools. The platform handles sandbox lifecycle, durable transcript storage, tool access, credential injection and workflow execution, then delivers progress updates and the final answer back into the same thread. It can also be driven directly through its API, and a lightweight local k3s cluster is enough for smaller setups, so no full production Kubernetes installation is required.
Centaur is open source and self-hosted: you run the platform on your own infrastructure and bring your own agent subscriptions or API access. There is no hosted paid tier from the project itself; your costs are your own servers and model usage.
Engineering and operations teams that want one shared, auditable agent for investigating CI failures, answering questions with internal tools, summarizing threads and running recurring checks, instead of scattered one-off local setups.