loading
Loading.Open source
We run AI agents in production on our own software before we touch yours. These are the tools we built to do it. All open source, all free to run yourself. No account, no upsell.
Mission control for a fleet of coding agents: persistent memory, inter-agent messaging, shared tasks, and one dashboard to watch them work.
Public, v1.7.2. Open source (AGPL).
How do you coordinate multiple AI agents? →Serves each coding agent the one current doc that answers a query (a path:line pointer with a freshness marker) instead of rereading the repo every session. About 60% fewer tokens per doc lookup, measured at equal task-success on our own repo (it varies by repo). Runs local: SQLite + ONNX, no cloud or keys.
Public beta, v0.13.0. Open source (AGPL core, MIT CLI).
How do AI agents remember context? →Know what your agents did overnight: their runs, decisions, and outcomes, in one place.
Public beta. Open source (AGPL).
What is fleet observability for AI agents? →Runs your settled, repeatable agent work as deterministic scripts in isolated containers, with no model in the loop. The mechanical 80% costs nothing to re-run, so your agent tokens stay on the 20% that needs judgment.
v0.4.3 beta. Open source (Apache-2.0).
How much do AI coding agents cost to run? →We don't sell “cheaper.” We sell prod-grade output from a team you already trust, and we run agents to get there. The suite is how we do it, and it's the proof we can. You can run all of it yourself, today, for free.
We build your product with agent fleets, or we turn your developers into the operators who run them, on your stack and your standards.
or read how the two doors work