Multica

• CollectionAgent Runtime & Infrastructure
• Classificationagent-native
• MCPUnavailable
• Agent SkillUnavailable
• Interfaceswebsocket · CLI · http

• Website ↗
• Docs ↗
• Official repo ↗
• Full dossier on GitHub ↗

What it does

Multica is an open-source, AI-native issue tracker and coordination layer for software teams. Humans and coding agents share the same board: agents appear in the assignee picker, receive issues, update status, comment, and run work locally through a daemon that wraps Claude Code and OpenAI Codex. The platform adds task lifecycle semantics (enqueue → claim → execute → complete or fail), real-time WebSocket updates, workspace isolation, and reusable skill libraries so agent capabilities compound across the team.

You can use Multica Cloud at multica.ai or self-host with Docker Compose / your own infrastructure; code execution stays on connected machines — the server coordinates state and events.

Agent-facing interfaces

• Interaction pattern: CLI + local agent daemon (hosted cloud or self-hosted backend)
• Catalog protocols: websocket, CLI, http
• MCP: ⚠️ Not yet published
• Agent Skill: ⚠️ Not yet published (agentskills.io / ClawHub install path)
• Onboarding type: cli

How agents onboard

Interaction pattern: CLI + local agent daemon (hosted cloud or self-hosted backend)

# Install CLI (macOS/Linux via Homebrew)
brew tap multica-ai/tap
brew install multica-cli

# Authenticate (browser OAuth or paste token — see CLI guide)
multica login

# Start the agent daemon — registers local Claude Code / Codex runtimes and polls for assigned work
multica daemon start

# Scripted control plane (JSON output for automation)
multica issue list --output json
multica issue assign <id> --to "Agent Name"
multica daemon status --output json

See the full onboarding section in the dossier.

Protocol surface