One-liner
Open-source platform that turns coding agents into real teammates. Hero: "Your next 10 hires won't be human." Sub: "Manage your human + agent workforce in one place." Vendor-neutral — works with Claude Code, Codex, Cursor Agent, OpenCode, OpenClaw, Hermes, Gemini, Pi, Kimi, Kiro CLI.
The category — a fourth lane
Multica is not a builder, not a runtime / sandbox. It sits above the coding agents and turns them into team-scale teammates with profiles, issues, comments, skill libraries, and a unified runtime panel. The closest analog in this vault is none — it's a new category, "agent orchestration."
Frame it as Linear-for-AI-coding-agents: humans + agents in the same workspace, assigning issues, leaving comments, reporting blockers, compounding skills. Tasks flow enqueue → claim → start → complete/fail with WebSocket-streamed live progress.
Multica is strategically a partner for CreateOS, not a competitor. They orchestrate agents → CreateOS executes them in sandboxes → MPP settles payment. Different layer of the stack.
ICP summary
Three concentric rings: 1. Solo developers running multiple coding agents — the daemon auto-detects local CLIs, plug-and-play. 2. Small engineering teams using AI agents on real projects — the "shared workspace + activity feed" framing is for this audience. 3. Self-hosting / regulated teams — Docker Compose / Kubernetes / single binary; "your data never leaves your network."
See Multica - ICP.
Feature inventory
- Multica - Agents as Teammates — agent profiles, board presence, assignee dropdown, autonomous comments.
- Multica - Reusable Skills — codified skill library (
SKILL.md+ steps); compound learning across the team. - Multica - Unified Runtimes — single dashboard for local daemons + cloud runtimes; usage charts, activity heatmaps, auto-detection.
- Multi-workspace isolation (each workspace has its own agents, issues, settings).
- Full task lifecycle:
enqueue → claim → start → complete/failwith WebSocket streaming. - CLI-first onboarding (
multica setup,multica daemon,multica issue create|list). - Desktop app (multi-tab native window, ships with CLI built-in).
UX flows captured
One screenshot in raw/Multica/image.png — board view (not yet itemized). Marketing copy describes board view, agent profile/assignee picker, task execution timeline, skills editor, runtimes panel.
full console UX walkthrough (settings nav, agent creation flow, skill authoring) — needs more screenshots.
Notable
- 21.3k GitHub stars, 2.6k forks — significant open-source momentum (per github.com/multica-ai/multica capture). 53 releases, latest v0.2.17 on 2026-04-26.
- Bilingual: English + 简体中文. Sustained Chinese localization is unusual.
- Vendor-neutral pitch: explicitly NOT tied to one agent backend. Lists 10 supported coding-agent CLIs.
- Self-host first, cloud second — Multica Cloud is currently waitlisted; self-host is the documented default.
- Tech stack: Next.js 16 (App Router) + Go (Chi, sqlc, gorilla/websocket) + PostgreSQL 17 with pgvector.
- Comparison they ship: Multica vs Paperclip — they explicitly position against another "agent simulator" tool and claim team + cloud-first + lighter governance as their angle.
Sources
- Multica - Source Links (
raw/Multica/links.md) - Multica — Project Management for Human + Agent Teams (homepage clip, 2026-04-28)
- multica-aimultica The open-source managed agents platform (GitHub README clip, 2026-04-28)
- Welcome (docs landing clip, 2026-04-28)
raw/Multica/image.png(board view screenshot)