{"slug": "aming-claw-zero-orchestration-multi-agent-coding", "title": "Aming Claw – Zero-orchestration multi-agent coding", "summary": "Aming Claw, a new open-source framework, enables multi-agent coding without requiring users to write orchestration code. The system uses an observer, such as a current Claude Code or Codex session, to manage a commit-bound project graph and assign workers to scoped files under isolated contracts, with replay tied to frozen commits rather than chat memory. The framework claims to be the first to combine zero-orchestration setup, graph-based scope decisions, and commit-bound contract replay for coding agents.", "body_md": "If you've used LangGraph supervisor, AutoGen GroupChat, or CrewAI for coding work, you've written the orchestration. Aming Claw asks for zero.\n\nThe observer is your current Claude Code or Codex session, not a new daemon.\n\nThe observer holds the project's commit-bound code graph. It decides which worker gets which files using two signals together: the requirement itself (LLM-side) and the code graph's structural boundaries: dependency, module, and function scope.\n\nEach worker runs under its own contract: scoped files, fence token, trace ledger, close gate. The full worker path runs each worker in an isolated git worktree against a frozen commit hash. The HN demo starts from your current Claude Code or Codex session as observer; scripted workers are a zero-setup fallback that uses the same contracts, fences, and replay logic.\n\nThe shared object is not the chat. It is not the workflow state. It is the project graph.\n\n```\n                  observer\n                     |\n           commit-bound project graph\n                     |\n        +------------+------------+\n        |                         |\n   Worker A contract         Worker B contract\n   scope A, fence A          scope B, fence B\n        |                         |\n      pass                 fail / interrupted\n        |                         |\n candidate diff A          replay B against X\n        |                         |\n        +------------+------------+\n                     |\n              ordered Git merge\n                     |\n          target graph reconcile once\n```\n\nThe case I want you to challenge:\n\n- Worker A and Worker B both receive contracts bound to commit hash X.\n- Worker A passes; its diff is accepted as candidate evidence.\n- Worker B fails mid-execution.\n- The observer replays Worker B against commit hash X. Worker B sees the original code, not Worker A's in-progress changes.\n- The replay passes, producing a clean diff against X — Worker B's contract scope and Worker A's contract scope are disjoint by design, so B's replay never touches files A already accepted.\n- Both accepted diffs land through an ordered Git merge.\n- The target project graph is reconciled once after the accepted change lands.\n- The backlog row closes only after the timeline and contract gates pass.\n\nWorker A and Worker B can both be Claude, both Codex, scripted local workers, or any compatible agent process. The coordination model is the same regardless of runtime.\n\nThe installed-user demo starts with your current Claude Code or Codex session as observer. Scripted workers are available for zero-setup reproducibility and CI, so you do not need two AI subscriptions to challenge the protocol. Live worker mode plugs in whichever AI runtime you have.\n\nWhat is not new: supervisors, handoffs, traces, shared workflow state, checkpoint replay, parallel branches. LangGraph has strong primitives for supervisors, state graphs, checkpointing, replay, and durable workflows.\n\nThe narrow claim: I have not found another open-source, plug-and-play coding-agent framework where:\n\n- the user writes zero orchestration code;\n- the observer decides scope from the project graph itself, not just the prompt;\n- workers run under commit-bound contracts with fenced files and trace ledgers;\n- replay is tied to the original contract and frozen commit instead of chat memory;\n- accepted work reconciles once against the target project graph before the next agent treats it as truth.\n\nIf you know one -- research prototypes, commercial products, open-source projects -- please send it to me. I'd genuinely like to know what to compare against.\n\nRepo: [https://github.com/amingclawdev/aming-claw](https://github.com/amingclawdev/aming-claw)\n\nHow to run the demo: [HN multi-agent challenge demo](/amingclawdev/aming-claw/blob/main/docs/hn-demo/README.md)\n\nMore cases, audit trails, and the design story: [Hope is not an engineering\ncontrol for AI coding agents](/amingclawdev/aming-claw/blob/main/docs/hn-demo/design-story.md)", "url": "https://wpnews.pro/news/aming-claw-zero-orchestration-multi-agent-coding", "canonical_source": "https://github.com/amingclawdev/aming-claw/blob/main/docs/hn-demo/article.md", "published_at": "2026-05-27 13:59:01+00:00", "updated_at": "2026-05-27 14:16:04.281209+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "ai-infrastructure", "large-language-models", "generative-ai"], "entities": ["Aming Claw", "LangGraph", "AutoGen", "CrewAI", "Claude Code", "Codex"], "alternates": {"html": "https://wpnews.pro/news/aming-claw-zero-orchestration-multi-agent-coding", "markdown": "https://wpnews.pro/news/aming-claw-zero-orchestration-multi-agent-coding.md", "text": "https://wpnews.pro/news/aming-claw-zero-orchestration-multi-agent-coding.txt", "jsonld": "https://wpnews.pro/news/aming-claw-zero-orchestration-multi-agent-coding.jsonld"}}