--- title: HF Hub Community Agent emoji: 🤗 colorFrom: yellow colorTo: red sdk: docker app_port: 7860 license: apache-2.0 short_description: Query HF users, orgs, discussions, PRs, and more --- # HF Hub Community Agent This Space runs a [fast-agent](https://fast-agent.ai/) agent specialized in querying Hugging Face Hub community features. ## Capabilities This agent can help you with: - **User & Organization Data**: Get user profiles, overview info, followers, and following lists - **Discussions & Pull Requests**: Browse, create, and manage discussions and PRs on repos - **Access Requests**: Manage access requests for gated repositories - **Collections**: List, view, and manage HF collections - **Likes & Interactions**: Query liked repositories and user activity ## Configuration The agent uses the HuggingFace API with: - `hf_api_tool.py`: Secure API tool with endpoint allowlist - `hf_hub_community.md`: Agent card with detailed method documentation ## OAuth Token Passthrough This Space is configured for **user-pays** inference: - Users provide their own HuggingFace token via `Authorization: Bearer ` header - Inference costs are billed to the user's HF account - Supports HF Inference API models ## Environment Variables Configure in Space Settings: - `HF_TOKEN`: Fallback token for unauthenticated requests (optional) - `HF_ENDPOINT`: Override HF endpoint (default: https://huggingface.co) - `HF_MAX_RESULTS`: Max results for list responses (default: 20) ## Usage This Space exposes a fast-agent MCP server. You can interact with it using: 1. **MCP Clients**: Connect using the Space URL with your HF token 2. **HTTP API**: Send requests with `Authorization: Bearer ` header 3. **Claude Desktop**: Configure as an MCP server ## Security The `hf_api_tool.py` includes: - Endpoint allowlist for security - Token validation and path traversal protection - Request-scoped token passthrough for OAuth ## Model Uses `gpt-oss` model (configurable via agent card frontmatter). ## More Info - [Fast Agent Documentation](https://fast-agent.ai/) - [Hugging Face API Documentation](https://huggingface.co/docs/huggingface_hub)