usa-spending -DEMO

USA Spending MCP Server is a Python-based proxy that exposes an MCP interface to interact with the USAspending.gov API. It is designed as a proof-of-concept tool for querying and working with USAspending data through the MCP protocol. You can run the server locally using either the uv tool or Poetry, and then wire it into Claude or another MCP consumer by providing the server path in the mcpServers configuration. Once running, the server will accept MCP requests and return responses sourced from USAspending.gov, enabling programmable access to spending data, awards, and related information.

To use the server with Claude, point Claude's MCP config to the running server path. If you installed via uv, use the installed executable path (for example, /path/to/usa-spending-mcp-server). If you installed via Poetry, locate the poetry-managed executable and reference that full path in Claude’s MCP config. The server itself supports standard MCP request/response flows, so you can issue queries, filters, and actions supported by the underlying USAspending API through the MCP interface.

The project includes formatting and testing guidance (Black, isort) and provides simple commands to run or update the server during development. You can start the server directly as usa-spending-mcp-server or via poetry run usa-spending-mcp-server, depending on your installation method, and then reconfigure Claude or any MCP consumer to target the server endpoint.

Prerequisites:

• Python 3.13+
• Poetry (https://python-poetry.org/) OR uv (https://docs.astral.sh/uv/)

Option A: Install via uv (Recommended for quick setup)

1. Install the tool via uv: uv tool install git+https://github.com/GSA-TTS/arc-usa-spending-mcp-server
2. Run the server: usa-spending-mcp-server

Option B: Development Setup with Poetry

1. Install dependencies: poetry install
2. Activate the virtual environment: poetry env activate
3. Run the server: poetry run usa-spending-mcp-server

Notes:

• The repository README suggests using either uv or Poetry depending on your workflow. After installation, you can obtain the exact executable path (for Claude integration) and update your MCP config accordingly.

Tips and common issues:

• The project is a proof-of-concept and not intended for production use.
• If you install via Poetry, the server executable is located within the Poetry-managed virtual environment; capture its full path for Claude integration.
• When updating the server in development, re-run installation steps (e.g., poetry install) to refresh dependencies.
• For Claude configuration, the MCP config should point to the server executable path and can include an empty env object if no environment variables are required.
• The server relies on Python 3.13+ and uses Black and isort for code formatting during development.