What it does

Multica wraps your existing coding agents—Claude Code, Codex, Cursor Agent, Kimi, and roughly a dozen others—in actual team infrastructure. You assign issues to agents through a normal board view, they execute on local or cloud runtimes via a daemon, and they report progress, blockers, and completions like any other teammate. There’s also a “Squad” layer where a leader agent routes work to the right member, so you can @FrontendTeam instead of guessing which individual agent is free.

The interesting bit

The name is a deliberate callback to Multics, the 1960s OS that invented time-sharing. The pitch: just as Multics let multiple users share one machine, Multica lets humans and agents share one project management system—complete with profiles, comment threads, and reusable “skills” that compound over time. It’s a bet that agent management will become as infrastructural as CI/CD.

Key highlights

• Vendor-neutral by design — works with 12+ agent CLIs, not locked to one provider
• Self-hostable — full server deploys via Docker; cloud option available
• Autopilots — cron/webhook-triggered recurring tasks that auto-create and route issues
• Real-time execution — WebSocket streaming of agent progress as tasks move through enqueue → claim → start → complete/fail
• Stack is conservative — Next.js frontend, Go backend (Chi, sqlc), PostgreSQL 17 with pgvector

Caveats

• The “reusable skills” feature is described aspirationally; the README doesn’t clarify how skills are extracted, versioned, or shared
• iOS mobile client exists but requires manual build to your own device—no App Store distribution mentioned
• GHCR image tags may not be published yet for some versions; fallback requires building from source

Verdict

Worth a look if you’re already running multiple agent CLIs and tired of context-switching between terminals and copy-pasted prompts. Skip it if you want a single-agent, chat-only workflow—this is explicitly for teams multiplexing humans and agents at scale.