Multica is an open-source managed agents platform written in Go that enables teams to treat coding agents as real teammates. The core concept transforms how teams work with AI by allowing users to assign tasks to agents just as they would to human colleagues, with agents autonomously picking up work, writing code, reporting blockers, and updating task statuses. The platform supports multiple agent providers including Claude Code, Codex, GitHub Copilot CLI, OpenClaw, OpenCode, Hermes, Gemini, Pi, Cursor Agent, Kimi, Kiro CLI, and Trae, making it vendor-neutral and flexible for different team preferences.
The platform's architecture consists of a Next.js 16 frontend with App Router, a Go backend using Chi router with sqlc and gorilla/websocket, PostgreSQL 17 with pgvector for the database, and a local daemon runtime that executes agent CLIs. The repository has grown to 39,042 stargazers as of the most recent tracking period and maintains active development with a median issue and pull request response latency of 0.0 hours and a mean latency of 16.2 hours across 1,358 tracked items. The most active contributor, Bohan-J, has logged 1,668 events, followed by forrestchang with 161 events and multica-eve with 109 events. The issue tracker shows 420 enhancement requests, 364 bug reports, and 20 questions as the most active label categories.
Key features include Agents as Teammates, where agents receive profiles, appear on project boards, post comments, create issues, and proactively report blockers. Squads allow grouping of agents and humans under a leader agent for stable routing as teams scale. Autonomous Execution manages the full task lifecycle from enqueue through completion with real-time progress streaming via WebSocket. Autopilots enable scheduling of recurring work through cron triggers, webhooks, or manual runs for tasks like daily standups and weekly reports. Reusable Skills allow solutions to become team-wide capabilities that compound over time. Unified Runtimes provide a single dashboard for all compute resources, including local daemons and cloud runtimes with auto-detection of available CLIs. Multi-Workspace support enables organization across teams with workspace-level isolation.
The platform is designed around the philosophy that agents are first-class teammates, with the assignee picker, activity timeline, task lifecycle, and runtime infrastructure all built with this principle from the start. Installation is straightforward through Homebrew on macOS and Linux, with a PowerShell option for Windows. Self-hosting is supported through Docker with official images from GHCR. The CLI provides commands for authentication, daemon management, workspace switching, and issue creation.
The repository is classified across multiple domains including Multimodal, Causal Inference, Causal Discovery, Causal Estimation, Data Integration, Machine Learning, AI, Explainability, Healthcare, and Finance. Related repositories with overlapping contributors include openclaw/openclaw, yeachan-heo/oh-my-claudecode, and nousresearch/hermes-agent. Development requires Node.js v20 or higher, pnpm v10.28 or higher, Go v1.26 or higher, and Docker, with a single make dev command handling environment setup, dependency installation, database configuration, migrations, and service startup. An iOS mobile client is also available in the apps/mobile directory.
