Aming Claw is a local AI governance workspace for turning a codebase into a queryable graph, then using that graph to guide AI-assisted issue discovery, backlog filing, and PR work.
The MVP is intentionally local-first: clone the repo, start the governance service, load the Codex or Claude Code plugin from the cloned checkout, then use the dashboard and MCP tools against your projects.
Requirements:
- Python 3.9+
- Git
- Node.js/npm when rebuilding the dashboard from source
- Codex or Claude Code for the plugin/skill workflow
If you are starting from a plugin host, give it the repository URL:
Install the Aming Claw plugin from https://github.com/amingclawdev/aming-claw
If the host cannot install Git plugins directly yet, clone once and use the repo-local package:
git clone https://github.com/amingclawdev/aming-claw.git
cd aming-claw
pip install -e .
aming-claw startIf the aming-claw console script is not on PATH yet, run the same CLI
through Python:
python -m agent.cli startOpen the local launcher or dashboard:
aming-claw launcher --open-browser
aming-claw openDashboard URL:
http://localhost:40000/dashboard
Aming Claw ships its plugin files inside this repository:
.codex-plugin/plugin.json.agents/plugins/marketplace.json.claude-plugin/plugin.json.claude-plugin/marketplace.json.mcp.jsonskills/aming-claw/skills/aming-claw-launcher/
Open the cloned aming-claw directory in Codex. The repo-local marketplace and
project .mcp.json provide the Aming Claw skill and MCP server contract.
In a new session, ask Codex to use the Aming Claw skill and check runtime state:
Use the Aming Claw skill. Check runtime_status, graph_status, and backlog before changing code.
Codex does not auto-start Aming Claw services. Start them explicitly:
aming-claw startIf the CLI is already available, it can update a local plugin checkout from the public Git URL:
aming-claw plugin install https://github.com/amingclawdev/aming-clawFrom Claude Code, add the Git repository as a marketplace and install the plugin:
/plugin marketplace add https://github.com/amingclawdev/aming-claw
/plugin install aming-claw@aming-claw-local
If your Claude Code build does not support Git marketplace URLs yet, clone the repo and add the local checkout instead:
git clone https://github.com/amingclawdev/aming-claw.git
cd aming-claw
pip install -e .
aming-claw start/plugin marketplace add /path/to/aming-claw
/plugin install aming-claw@aming-claw-local
The Claude plugin exposes the main governance skill and the launcher skill:
/aming-claw:aming-claw/aming-claw:aming-claw-launcher
- Clone and install Aming Claw.
- Start the local governance service.
- Load the Codex or Claude Code plugin.
- Bootstrap or select a project in the dashboard.
- Build or update the project graph.
- Inspect nodes, files, functions, docs, tests, and relations.
- Use AI Enrich on selected nodes or edges.
- Review and accept or reject proposed semantic memory.
- File backlog rows and PR opportunities with graph evidence.
Register projects, switch active workspaces, configure AI routing, and trigger graph builds or updates.
Explore the project structure, dependency relations, health, files, tests, docs, and function indexes.
AI Enrich generates semantic summaries, intent, risk, and evidence for selected nodes or edges. Generated semantics are not trusted memory until reviewed by an operator.
Use selected-node or selected-edge enrichment first on large projects to control cost and review scope.
Typical AI Enrich flow:
- Select a graph node or edge.
- Click
AI enrich. - Watch the job in Operations Queue.
- Review the proposed semantic memory.
- Accept, reject, or retry.
- Use accepted semantics for backlog and PR planning.
Track graph updates, semantic jobs, review work, and background operations.
Review AI-generated semantic proposals before they become trusted project memory.
aming-claw launcher # write/open the local launcher
aming-claw plugin install # clone/update local plugin assets from Git
aming-claw start # start governance locally
aming-claw open # open the dashboard
aming-claw status # check governance health
aming-claw bootstrap # register a project workspace
aming-claw scan # scan an external projectFor MVP distribution, use Git rather than PyPI:
git clone https://github.com/amingclawdev/aming-claw.git
cd aming-claw
pip install -e .For an already-installed Aming Claw CLI, refresh a user-local plugin checkout from Git:
aming-claw plugin install https://github.com/amingclawdev/aming-clawFor a cloned checkout before the console script is on PATH, run the installer
through Python:
python scripts/install_from_git.py https://github.com/amingclawdev/aming-clawPython-only installation from Git is also possible:
pip install git+https://github.com/amingclawdev/aming-claw.gitThe Python-only path installs CLI entrypoints, but local plugin loading is
clearest from a cloned checkout because Codex and Claude Code need access to the
repo-local plugin manifests, skills, and .mcp.json.
Before publishing a release build, rebuild the dashboard assets:
npm --prefix frontend/dashboard run build
python scripts/build_package.py --skip-dashboard-buildWhen working on Aming Claw itself, follow the project governance contract:
- Check runtime, graph, and queue state before implementation.
- File or update a backlog row before mutating code, docs, config, dashboard assets, or runtime state.
- Query the graph before creating new modules or changing behavior.
- Follow
skills/aming-claw/references/mf-sop.mdfor manual fixes. - Evaluate whether E2E evidence is needed for dashboard or graph behavior.
- Architecture
- Deployment
- Governance Overview
- Configuration Reference
- API Overview
- Plugin Packaging Notes
Aming Claw is licensed under the Functional Source License, Version 1.1, MIT Future License (FSL-1.1-MIT). See LICENSE.