Hi Product Hunt 👋 I built TeammateX because joining a large codebase is still brutal. Most AI tooling reads files. It does not know who wrote them, why they exist or what breaks when you touch them. So I gave the agent a memory. TeammateX clones your repositories and mines real git history into a knowledge graph. Files, functions, classes, call edges and human ownership become nodes and relationships in Neo4j, sitting alongside code embeddings in pgvector. That pairing is the whole idea. Vector search guesses by meaning. The graph knows by structure. The result is a teammate that can tell you the retry logic lives in queue/consumer.py line 142, that John owns it, that it is called from three places and which one of them can deadlock. Every claim is backed by a lookup you can verify. What I care about most: Self-hosted first. One Docker Compose file on your own VM. Nothing leaves the box unless you send it to a model, and you can remove even that by running Ollama locally. Real product surfaces. Standup, tasks, costs and audit are deterministic pages backed by real tables, not an LLM guessing. Honest cost tracking. Every model call is metered, not only chat, so the dashboard shows true spend with an optional monthly budget cap. It ships with 28 agent tools and it runs git and gh itself, so it can branch, commit and open a real pull request rather than describing one. It is alpha and MIT licensed. It runs end to end today but expect rough edges. I would rather ship early and learn in the open than polish alone. Tell me what breaks and which repository you point it at first. I will be here all day. Cheers and Thank you! ❤️

teammate X

AI teammate for your entire organisation's codebase

AI teammate for your entire organisation's codebase