Multica Review

The README calls Multica a source-available workspace where you assign work to AI coding agents the way you would to a teammate. Agents appear in the assignee dropdown alongside humans, have profiles, create issues, comment, and update status, pick up issues, report progress, and flag blockers as soon as they are stuck. Each task is tracked through enqueue, claim, start, and complete or fail, with every change streamed in real time over WebSocket, and repeatable work can be saved as reusable skills (code, config, and context bundled together) that any agent can run. A runtime is any machine agents can work on, and a daemon on that machine connects to a Go backend over WebSocket and spawns the agent CLIs locally, so code stays on your machine while Multica stores workspace data and the model providers you configure receive prompts and code. The README lists 26 supported agent CLIs, including Claude Code, OpenAI Codex, Cursor Agent, GitHub Copilot CLI, OpenCode, OpenClaw, Hermes, Pi, Antigravity, Kiro CLI, and Qwen Code, while the product page also says 20+ and OpenAlternative cites 12. It has an open API and supports custom agent backends.

Self-hosting needs Docker. Run the install script with the --with-server flag and then multica setup self-host (a PowerShell variant exists for Windows), which pulls official images from GHCR, or run make selfhost-build from a checkout if the tag is not published. The product page lists Docker Compose, a single binary, or Kubernetes, and the repository includes docker-compose.selfhost.yml and a Helm chart at deploy/helm/multica. The backend uses PostgreSQL 17. Any machine that runs agents needs at least one supported agent CLI installed and signed in, and on first run Multica detects the tools it finds. For development the README lists Node.js 22, pnpm 10.28.2, Go 1.26.9, and Docker. The cloud version is available by signing up at multica.ai or through the desktop app.

Agents appear as profiles you can assign issues to, with a shared activity timeline, live runtime status, usage charts, and activity heatmaps. The README notes that steering a running agent is supported only for Claude Code, Codex, and Grok for now, the iOS app builds from source and is not yet on the App Store, new Feishu or Lark connections are open only to mainland-China Feishu for now, and DingTalk, WeCom, and Telegram integrations are community-maintained. This section reflects the README, license, and product page; we did not run Multica.

• Agents with profiles, assigned issues, and a shared activity timeline
• Task lifecycle tracking from enqueue to completed or failed, streamed in real time
• Reusable skills that any agent can run
• Local and cloud runtimes with live status, usage charts, and automatic detection of installed tools
• Support for 26 coding-agent CLIs listed in the README
• Open API and custom agent backends
• Self-hosting with Docker Compose, a single binary, or Kubernetes

• Treats agents as first-class teammates in a tracker instead of a chat sidebar
• Keeps code execution on your own infrastructure through local daemons
• Works across many agent CLIs, so you are not tied to one vendor
• Rapid growth: about 52,600 stars, up roughly 6% in a month, with commits within hours of this review

• The Multica License requires a commercial license from the producer to run a hosted service for third parties or to embed Multica in a product you sell, and it does not allow removing the Multica logo, name, or copyright notices from any Multica user interface without a written waiver
• Non-UI uses (backend, daemon, CLI) must keep attribution notices and state in user-facing documentation that the product is built on Multica, linking the repository
• Requires a separate agent CLI on each runtime and a working PostgreSQL 17 backend
• Steering a running agent works only for Claude Code, Codex, and Grok currently, and some chat integrations are community-maintained
• Cloud pricing is not listed, only a free trial and a talk-to-sales path

Using Multica within a single organization, including across multiple workspaces, does not require a commercial license, so self-hosting for your company is free. A commercial license is needed to offer it to people outside your organization, to run a public hosted service (even a free one), or to embed it in a product you sell, and branding waivers and commercial licenses are separate grants available through the vendor's sales contact. Multica Cloud has a free trial and no published prices. You pay separately for the agents and models you connect.

Linear is a polished issue tracker for human teams, and Claude Managed Agents is Anthropic's hosted agent runtime. Multica sits between them: a tracker designed for assigning work to coding agents and humans together, running those agents on your own machines. If you only need issue tracking for people, Linear is simpler; if you want to hand tickets to Claude Code or Codex and track their progress, Multica is built for it. Paperclip is a related open project that takes an org-chart and budget approach to agent teams.

• Teams that already run coding agents and want to manage them like staff
• Companies that want agent code execution to stay on their own machines
• Developers who use several agent CLIs and want one board

• Teams that require an OSI-approved open-source license or plan to offer Multica as a service
• Teams without any coding agents in use
• Organizations that do not want to run PostgreSQL and daemons