ghx is a GitHub execution router for AI agents -- one typed capability interface over gh CLI and GraphQL. It validates input, selects the optimal route, handles retries, and returns a stable ResultEnvelope.
Start here: Root README for quick start and overview, or Core Getting Started for detailed setup.
%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#4A90D9', 'primaryTextColor': '#fff', 'primaryBorderColor': '#2E6BA4', 'lineColor': '#666', 'fontSize': '13px'}}}%%
graph TD
Hub["Documentation Hub"]
Hub --> Core["@ghx-dev/core"]
Hub --> Profiler["@ghx-dev/agent-profiler"]
Hub --> Eval["@ghx-dev/eval"]
Hub --> Repo["Repo-Wide"]
click Core "../packages/core/docs/README.md"
click Profiler "../packages/agent-profiler/docs/README.md"
click Eval "../packages/eval/docs/README.md"
style Hub fill:#4A90D9,color:#fff
style Core fill:#9C27B0,color:#fff
style Profiler fill:#F5A623,color:#fff
style Eval fill:#E91E63,color:#fff
style Repo fill:#7ED321,color:#000
| Package | Description | Docs |
|---|---|---|
@ghx-dev/core |
Public npm package -- CLI + capability routing engine | packages/core/docs/ |
@ghx-dev/agent-profiler |
Generic AI agent session profiler | packages/agent-profiler/docs/ |
@ghx-dev/eval |
Evaluation harness for ghx benchmarking | packages/eval/docs/ |
- Evaluation Report -- Empirical evaluation, statistical analysis, and bundled raw data
- Repository Structure -- Monorepo layout and module organization
- Architecture -- Package structure, execution flow, and design overview
- Troubleshooting -- Common setup, runtime, and CI issues with fixes
- Contributing -- Development setup, testing, CI, publishing
- Roadmap -- Current priorities and capability batches
- Blog Post: AI Agents Shouldn't Relearn GitHub on Every Run -- Full motivation and benchmark methodology