gh extension install github/gh-aw
Start Here
Use the table as a quick chooser. The goal is to separate agent workflow execution, model experimentation, activity summarization, and MCP setup.
Starter Pack
Inspect the workflow-specific install bundle, then install only the extensions that match your workflow. Do not pipe install bundles directly into a shell.
curl -fsSL https://sjh9714.github.io/gh-extension-atlas/install/starter-packs/ai-and-agents.txt
gh extension install github/gh-aw
gh extension install github/gh-models
gh extension install shuymn/gh-mcp
| If you need to... | Start with | Why | Status | Install |
|---|---|---|---|---|
| Run GitHub-native agent workflows | aw Run GitHub agentic workflows from the CLI. |
The strongest first stop when the workflow itself is an agent task. | active | gh extension install github/gh-aw |
| Test prompts and models from the terminal | models Use GitHub Models from GitHub CLI. |
Useful when you want GitHub Models inside an existing gh workflow. | active | gh extension install github/gh-models |
| Generate an AI-assisted standup report | standup Generate an AI-assisted standup report from GitHub activity. |
A focused helper when the repeated task is summarizing GitHub activity. | active | gh extension install sgoedecke/gh-standup |
| Connect AI tools to GitHub through MCP | mcp Run GitHub MCP Server using existing GitHub CLI authentication. |
A practical bridge for running GitHub MCP Server with existing gh authentication. | active | gh extension install shuymn/gh-mcp |
First Picks
gh extension install github/gh-models
gh extension install shuymn/gh-mcp
AI Workflow Fit
Use agent workflow tools when GitHub is the work surface, Models tools when prompt iteration is the task, and MCP helpers when another AI client needs authenticated GitHub context.
gh extension install sgoedecke/gh-standup
Freshness Notes
The atlas is a reviewed snapshot, not a live ranking. Recheck upstream repositories before adopting an AI or agent extension, especially when the extension can access repository data, call external models, or configure MCP access.