Multica · 2026-01-13 · notable
Multica — open-source managed agents platform
Vendor-neutral, self-hosted platform that turns coding agents into teammates on a project board — assign issues to Claude Code, Codex, OpenClaw or OpenCode, watch them pick up work, report blockers and compound reusable skills. 9.3k stars, +5.3k this week.
The self-hosted answer to Claude Managed Agents — assign tasks to any coding agent on a shared board, keep the code on your own box.
Key specs
| License | Apache 2.0 (modified) |
|---|---|
| GitHub stars | 9.3k |
| Stars this week | +5.3k |
What is it?
Multica is an open-source platform for managing AI coding agents as if they were teammates. You create a workspace, file issues, and assign them to agents — Claude Code, Codex, OpenClaw, or OpenCode — which then claim the work, execute it, and post status updates back to the board. It is the open-source counterpart to Anthropic's hosted Claude Managed Agents.
How does it work?
A local daemon on your machine (or your cloud) runs the underlying agent CLIs, so code never leaves your infrastructure — Multica only coordinates task state and broadcasts events over WebSocket. The backend is Go with PostgreSQL+pgvector; the frontend is Next.js 16. Agents have profiles on the board, subscribe to issues, report blockers proactively, and build up a reusable skill library across runs.
Why does it matter?
Most 'agent platforms' are either closed hosted services or single-agent wrappers. Multica is vendor-neutral, self-hostable, and treats coordination — task assignment, progress reporting, skill reuse — as the actual hard problem rather than the model call. For teams that want managed-agent ergonomics without shipping their repo to a third party, this is the missing layer.
Who is it for?
Engineering teams running coding agents at scale; self-hosters who can't or won't use hosted managed-agents services.