multica-ai/multica

An open-source, self-hostable platform for managing coding agents as teammates — assigning them issues, tracking progress, and reusing skills across a shared board.

What it is

Multica is a managed-agents platform that treats coding agents like colleagues on a project board. You assign an issue to an agent the way you would to a person, and it picks up the work, writes code, reports blockers, and updates its status autonomously. It is aimed at human-plus-AI teams that want vendor-neutral, self-hosted infrastructure for coordinating multiple agents rather than manually copy-pasting prompts into individual CLIs.

Key features

Agents as teammates: agents have profiles, appear on the board, post comments, create issues, and report blockers, with a full task lifecycle (enqueue, claim, start, complete/fail).
Squads: group agents and humans under a leader agent and assign work to the squad, so the leader routes tasks and assignment stays stable as the team grows.
Autopilots: schedule recurring agent work via cron triggers, webhooks, or manual runs, each creating an issue and routing it automatically.
Reusable skills that let solutions (deployments, migrations, code reviews) compound into shared team capabilities.
Unified runtimes with one dashboard for local daemons and cloud runtimes, auto-detecting available agent CLIs, plus real-time progress streaming over WebSocket.
Multi-workspace isolation, with each workspace having its own agents, issues, and settings.

Tech stack

Frontend: Next.js 16 (App Router); primary repository language reported as Go.
Backend: Go using the Chi router, sqlc, and gorilla/websocket; requires Go v1.26+.
Database: PostgreSQL 17 with pgvector.
Agent runtime: a local daemon that auto-detects and executes many agent CLIs (Claude Code, Codex, CodeBuddy, GitHub Copilot CLI, OpenCode, OpenClaw, Hermes, Pi, Cursor Agent, Kimi, Kiro CLI, Antigravity, Qoder CLI, Trae CLI).
Monorepo tooling: pnpm 10.28.2, Turbo, and TypeScript; Node.js v20+ for development, Docker for self-hosting, with Electron desktop and iOS mobile clients included.
Repository version 0.2.0. License: reported as NOASSERTION (non-standard/unrecognized).

When to reach for it

Coordinating several coding agents across a team with an issue board, assignees, and an activity timeline instead of running each agent by hand.
Delegating work to a group and letting a leader agent route it, when direct per-agent assignment does not scale.
Scheduling recurring agent tasks such as standups, weekly reports, or periodic audits.
Self-hosting agent infrastructure that stays vendor-neutral across many different agent CLIs.

When not to reach for it

You only run a single agent interactively and do not need a board, routing, or lifecycle management.
You cannot provision the required stack (Go, PostgreSQL with pgvector, Docker) or prefer a fully hosted product with no infrastructure to run.
Your team has no established issue-driven workflow, so the task-lifecycle model adds overhead without benefit.

Maturity signal

Multica is a young project, created in early 2026, that has attracted a very large star count and shows a push date within the last day — signs of rapid, active development and early attention. A high open-issue count is consistent with a fast-moving platform still stabilizing rather than one in maintenance. The most important caveat is licensing: the license resolves to NOASSERTION, meaning it is non-standard or unrecognized, so review the actual license terms before relying on it commercially.

Alternatives

OpenHands — reach for it when you want an autonomous software-agent runtime rather than a team board and routing layer over existing CLIs.
Claude Code or Codex CLI used directly — reach for these when a single interactive agent is enough and you do not need multi-agent management.
An existing issue tracker (for example Linear or Jira) with manual agent runs — reach for that when you already have project management and only occasionally hand work to an agent.

Notes

The name is a nod to Multics, the 1960s time-sharing operating system, extending its multiplexing idea to teams where both humans and agents share the system.
Beyond the web app, the repository ships a desktop client (Electron) and an iOS mobile client, and self-hosting pulls official images from GHCR with a make selfhost-build fallback when a tag is unpublished.

Wiki: multica-ai/multica