- Category
- AI Agents
- Rank
- No. 97Tools index
- Pricing
- Open Source
- Type
- APP
- Builder
- @jiayuan_jy
- GitHub
- 49.2k stars
- Latest release
- v0.4.41
- Date
About
An open-source platform that turns AI coding agents like Claude Code and Codex into real teammates. Assign tasks to agents like human colleagues, track their progress on a board, and let them autonomously write code, report blockers, and compound skills over time.
What it does
Multica gives agent runs a shared operational record. Issues hold the request, ongoing comments, execution history, decisions, usage, errors, and resulting changes. Agents execute through authenticated command-line tools on connected machines, while people supervise work through web, desktop, mobile, chat, CLI, or API surfaces.
Why it's ranked here
The case is strong for teams already juggling several coding agents. Multica combines broad runtime support, controlled local execution, review gates, retries, cost visibility, scheduled work, and self-hosting. Its value depends on existing agent subscriptions and infrastructure, so it is an orchestration layer rather than a complete coding system.
What's good
Runs remain attached to their originating issues, including commands, errors, decisions, usage, and diffs. Human approval precedes shipping. Teams can reuse solved procedures as skills, schedule recurring work, restrict which agents members may run, and connect several Git hosts. Local runtimes keep code on machines the operator controls.
Tradeoffs
Multica supplies neither models nor agent command-line tools. Every runtime machine needs a supported tool installed, authenticated, and maintained separately. Full self-hosting adds Docker, PostgreSQL, configuration, and deployment work. Mobile support is currently iPhone-focused, and its iOS build must be compiled from source rather than installed from the App Store.
How to use it well
Use it when an engineering team runs multiple agent providers and needs one queue, durable context, auditability, cost tracking, and explicit review. Start with bounded issues on a controlled runtime, then capture repeated solutions as skills and automate stable routines. It does not replace an agent provider, model access, source control, or human code review.
Technical notes+
README.md describes a Next.js 16 frontend, Electron desktop client, Expo mobile client, Go backend using Chi and WebSockets, PostgreSQL with pgvector, and a local daemon that spawns agent CLIs. Makefile provides development, verification, database, daemon, CLI, Docker Compose, and worktree workflows, with safeguards against remote database resets and legacy Compose. package.json defines a private pnpm and Turborepo workspace with TypeScript, Playwright, Electron, and filtered mobile builds. packages/core/api/index.ts implements shared HTTP and WebSocket clients through a boot-time singleton; packages/core/auth/index.ts and packages/core/chat/index.ts use registered Zustand store proxies. packages/core/i18n/index.ts separates server-safe, React, and browser entry points to avoid DOM and React context failures in server environments.
Observed
- License
- Described as open source; the supplied repository text does not identify a specific license.
- Primary languages
- Go powers the backend, daemon, and CLI; TypeScript powers the Next.js, Electron, and Expo clients.
- Packaging and install
- Offers hosted signup, desktop downloads, shell and PowerShell installers, Docker Compose self-hosting, Helm self-hosting, and source development through pnpm, Turborepo, Go, and Make.
- Interfaces
- Web, desktop, iPhone, CLI, API, WebSocket, and chat-channel surfaces are documented.
- Platform support
- Desktop supports macOS, Windows, and Linux; iOS builds are available from source.
- Runtime support
- Drives twenty separately installed and authenticated agent CLIs through a local daemon.
- Storage architecture
- Uses PostgreSQL with pgvector behind a Go backend.
- Verification surface
- The Make verification workflow includes TypeScript checks, TypeScript tests, Go tests, and Playwright end-to-end tests.
Read from README.md, Makefile, package.json, packages/core/index.ts, packages/core/api/index.ts, packages/core/vcs/index.ts, packages/core/auth/index.ts, packages/core/chat/index.ts, packages/core/i18n/index.ts, packages/core/lark/index.ts, packages/core/pins/index.ts, packages/core/inbox/index.ts, packages/core/paths/index.ts, packages/core/slack/index.ts, packages/core/types/index.ts.
What it can do
Assign coding tasks to AI agents
Task descriptions and requirements → Task assignments distributed to AI coding agents
Track AI agent progress on task board
Ongoing agent activities and task status → Visual progress board showing task completion status
Generate code autonomously via AI agents
Code requirements and specifications → Written code files and implementations
Report development blockers automatically
Agent execution attempts and encountered issues → Blocker reports and issue notifications
Build reusable skills for AI agents
Completed tasks and successful code patterns → Accumulated agent skills and capabilities
Manage multiple development workspaces
Project requirements and workspace configurations → Organized multi-workspace development environment
Integrate AI agents into development workflows
Existing development processes and AI coding agents → Integrated workflow with embedded AI agents
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.