aura

Aura is a Python-based MCP server that ingests and exposes Finnish open data as a single, searchable interface for AI models and humans alike. It aggregates metadata from 28 data sources, normalizes formats such as CKAN, PxWeb, OpenAPI, WFS, and more, and makes datasets searchable with natural language. The server also provides quality scoring, schema inference, and enrichment capabilities, enabling AI agents (like Claude or GPT) to access rich, structured dataset information through MCP protocol endpoints. Core tools include natural language search, structured search for AI-ready responses, region-based queries, dataset descriptions, and data preview/probing across multiple formats. You can use Aura directly or integrate it into tools that support MCP. Tools listed under Haku ja selaus, Alueanalyysi, Laatu, Rikastus ja tutkimus, and Hallinta provide a broad set of capabilities for discovery, quality assessment, enrichment, and management of the data catalog.

To use Aura as an MCP server, run the standard Python command and connect clients that speak the MCP protocol. Depending on your workflow, you can interact with Aura through the CLI (aura search, aura stats, aura harvest, aura infer-schemas, etc.) or configure client tools like Claude Desktop, Cursor, Windsurf, or any MCP-compatible tool to point at the aura server.

Prerequisites:\n- Python 3.11+ installed on the system (check with python3 --version)\n- Git installed to clone the repository\n- pip available for Python package installation\n\nStep-by-step installation:\n1) Clone the Aura repository:\n\nbash\ngit clone https://github.com/trotor/aura.git\ncd aura\n\n2) Create and activate a Python virtual environment:\n\nbash\npython3 -m venv .venv\nsource .venv/bin/activate # macOS/Linux\n# On Windows: .venv\Scripts\activate\n\n3) Install the package in editable mode so updates reflect immediately:\n\nbash\npip install -e .\n\n4) Run the MCP server (Aura) locally:\n\nbash\npython -m aura.cli serve\n\n5) Optional: pre-install optional dependencies or extra tools per your environment as described in the README (e.g., Claude Code/Claude Desktop setup, Cursor, Windsurf).\n\nNotes:\n- The repository ships with a prepopulated data store (data/aura.db); harvest is not required for basic operation.\n- Ensure your Python virtual environment is activated whenever running Aura.\n

Tips and considerations:\n- Aura exposes an MCP server named 'aura'. You can configure clients to point to this server using the standard MCP tooling conventions.\n- The README shows multiple integration routes: Claude Code/Claude Desktop, Cursor, Windsurf, and generic MCP-compatible tools. You can run Aura as a standard MCP server with the command: /path/to/venv/bin/python -m aura.cli serve.\n- If you plan to deploy in an environment without Python installed globally, consider using uv (Python/uv) or packaging approaches that fit your orchestration strategy.\n- Environment variables and config can be extended via the existing .mcp.json when using Claude Desktop, Cursor, Windsurf, etc. The Aura server will read the .mcp.json to auto-start when appropriate.\n- For data sources, Aura aggregates data from 28 sources and supports formats like CKAN, PxWeb, OData, and WFS; use aura harvest and aura infer-schemas to refresh metadata and schemas as needed.\n- If you extend Aura with an API key (e.g., MML API), include it in your environment or config to access restricted data sources.