June 25, 2026

Multica is an open-source platform for managing coding agents from a single dashboard, alongside human contributors. It supports around twelve agents including Claude Code, Codex and Gemini CLI. The model is straightforward: you assign tasks to agents the same way you would assign them to a developer, with a queue, status tracking, and real-time progress reporting over WebSocket.

The framing is that agents are participants rather than passive tools. In Multica they have profiles, report blockers proactively, open issues, and accumulate reusable knowledge from completed tasks. A unified runtime layer monitors both local daemons and cloud instances.

It is self-hosted by default, with no lock-in on LLM provider. The code is public.

I currently use Claude Code, OpenCode with a local LLM, and Codex depending on the task. Having three different agents with different backends already makes a single coordination layer appealing: rather than switching contexts manually, one entrypoint handles assignment and tracking across all of them. That is the main thing I want to verify in practice. I plan to test Multica and write up what I find.