DEV Community

정상록
정상록

Posted on

Multica: Managing AI Coding Agents as Real Teammates on a Kanban Board

#ai #agents #opensource #productivity

TL;DR

Multica is an open-source agent management platform where AI coding agents (Claude Code, Codex, OpenClaw, OpenCode) sit alongside human teammates on a Kanban board. ~3,900 GitHub stars in 3 months. Apache 2.0.

The Problem: Agent Information Silos

If you're using AI coding agents, you've probably hit this wall: one agent solves a problem, but the rest of your team has no idea it happened.

There's no shared visibility. No task tracking. No way to see what Agent A is doing while Agent B works on something else.

What Multica Does

Instead of building another coding agent, Multica provides a management layer for existing ones.

Your existing agents     Multica           Your team
┌──────────────┐    ┌──────────────┐    ┌──────────┐
│ Claude Code  │───▶│              │◀───│ Human 1  │
│ Codex        │───▶│  Kanban Board│◀───│ Human 2  │
│ OpenClaw     │───▶│  + WebSocket │◀───│ Human 3  │
│ OpenCode     │───▶│              │    └──────────┘
└──────────────┘    └──────────────┘
Enter fullscreen mode Exit fullscreen mode

Key Features

1. Agents as Teammates

Assign profiles, names, emojis to agents. Drop tasks onto them via Kanban. They autonomously comment, create issues, and report blockers. Trigger with @mention.

2. Reusable Skills

This is the game-changer. When one agent learns deployment, that skill becomes available to ALL agents:

Day 1:  Teach Agent-A to deploy
Day 30: All agents deploy, test, and review code
Enter fullscreen mode Exit fullscreen mode

Import skills from ClawHub and Skills.sh.

3. Unified Runtimes

Local daemon + cloud runtimes in one dashboard. Auto-detects installed CLIs. Per-task token usage tracking (v0.1.9).

4. Real-time Everything

WebSocket streaming for live progress. Multiple agents can work on the same issue simultaneously.

Tech Stack

Layer Tech
Frontend Next.js 16 (App Router)
Backend Go (Chi, sqlc, gorilla/websocket)
Database PostgreSQL 17 + pgvector
Runtime Local daemon executing agent CLIs

Quick Start

Cloud (fastest):
Visit multica.ai — zero setup.

Self-host:

git clone https://github.com/multica-ai/multica.git
cd multica
cp .env.example .env
docker compose up -d
cd server && go run ./cmd/migrate up && cd ..
make start
Enter fullscreen mode Exit fullscreen mode

CLI:

brew tap multica-ai/tap
brew install multica
multica login
multica daemon start
Enter fullscreen mode Exit fullscreen mode

Recent Updates (v0.1.9)

  • Sub-issues support
  • Multiple agents on same issue
  • Per-task token usage tracking
  • Google OAuth login
  • OpenClaw/OpenCode runtime support
  • iOS app and Windows CLI (PRs in progress)

Who Is This For?

  • Small (1-10 person) AI-native teams
  • Teams already using Claude Code or Codex heavily
  • Anyone who needs visibility into what their AI agents are doing

Caveat: Still v0.1.x — impressive velocity (22 releases in 3 months) but production stability needs further validation.

Links

Top comments (0)