Open Source Radar — October 30: OpenMausBot, an Open Grok Bot
Today's board keeps circling one question: what if your coding agent had coworkers? Two projects answer with a desktop app that hands every bot its own computer; a third gives a single session the org chart of a game studio.
OpenMausBot (TypeScript, 7,848 stars, Apache-2.0) is the open-source answer to xAI's Grok Bot — a chat app where each contact in the sidebar is a real agent, Claude or Codex running under the hood, with its own personality, its own model, its own sandboxed computer and its own connected apps. You talk to them like contacts, watch them work, and approve what actually matters. The project gained 1,199 stars today and 1,143 forks, and it is explicit that it is independent: no affiliation with xAI, and the same substitution play aimed at Meta Muse, OpenAI dots and Cue by Manus. The bet worth noting is isolation over convenience — one bot misbehaving does not take your shell with it.
OpenBot (TypeScript, 4,120 stars) from nightly-labs is the other end of the same idea, and the name collision with the CopilotKit project we covered yesterday is not the only reason to read the licence. Every agent gets its own workspace and conversation, plus local queues, file transfers, an embedded browser and agent-to-agent messaging, across Codex, Claude Code, Grok CLI, OpenCode, Gemini, Cursor and Cline. It picked up 709 stars today. Two caveats the README states plainly: it is a development preview where agents run with full filesystem and network access and no per-action approvals after first-launch consent, and it ships under PolyForm Noncommercial — source-available, not open source. If you plan to build a product on it, that distinction is the whole decision.
Claude Code Game Studios (Shell, 29,846 stars, MIT) turns one session into 49 agents and 74 skills arranged like a real studio: directors guarding the vision, department leads owning their domains, specialists doing the work, with 15 hooks and 14 rules holding the line. The pitch is specific and correct — a single chat has nothing stopping you from hardcoding magic numbers, skipping the design doc or writing spaghetti, and nobody asks whether a feature fits the vision. (repo). It resurfaced with 353 stars today despite last being pushed October 19, and 4,092 forks suggest people are already forking the hierarchy rather than just installing it. Worth it for a solo dev shipping on Godot, Unity or Unreal; the token bill for 49 agents is the part nobody budgets.
AI Design Skills (MIT, 12,624 stars) is the smallest thing on this list and the most revealing: a collection of design skills for coding agents that currently contains exactly one, landing-page-design — intake questions, page structure, conversion copy, and a full visual system covering type, spacing, radius and motion, with every design value in one editable section near the bottom so a fork becomes your house style. (repo). It pulled 1,187 stars today and 917 forks, which is a strange ratio for a single skill: people are not collecting it, they are changing it. Treat it as evidence that agents with a taste file beat agents with a blank prompt — not as a library yet.
Worth watching this week: whether per-bot sandboxes become the default shape for these products, and how many more PolyForm-licensed agents get listed as open source before the label stops meaning anything.
If you build with any of these, tell us what broke first — especially the licence question. Write to us via our contact page.
