federal_accountsLABSCO SUMMARY
Federal spending questions, answered against the government's own figures.
The tools follow the way the question is usually asked — how much, to whom, through which account, and how it changed — so 'what does the Department of Education spend on staff' resolves in a call or two instead of a session with the API docs open.
What it is
A server over the USASpending.gov API — the official record of federal awards and agency budgets — running as a stateless streamable-HTTP endpoint.
What you get
- Spending over time for a period you choose
- Award-level spending, filtered down to the subset you care about
- The federal accounts an agency receives and spends its congressional funding through
- Every budget function, ordered by title, so one policy area can be set against another
- Major object class breakdowns — what an agency spends on pay and benefits versus everything else
- The list of top-tier agencies, and the total budgetary resources behind them
Requirements
No account and no key — USASpending.gov is open data. Run it locally with uv, from PyPI with uvx, or from the Docker image on port 8000; the host and port it binds to are both overridable through the environment. Clients without streamable HTTP reach it through mcp-remote.
Setup effort
One command — uvx --from usaspending-mcp-server@latest usaspending-mcp-server
USE CASES
Tracing how one agency's money actually moves
Read the federal accounts the agency draws on, then break its spending down by object class to see what the money is spent on.
Showing how a spending line has moved
Pull the same figure across a date range rather than comparing two annual reports by hand.
Finding who received the money
Drill into award-level records for a period and filter to the recipients or programs in question.
Comparing two policy areas
Budget functions give the comparable buckets — community and regional development against international affairs, for example — with the totals behind them.
Getting your bearings on which agency is which
List the top-tier agencies first, so later calls name the agency the API recognises.
10 tools
Listed in project docs. Your client may show a different list.
list_budget_functionsmajor_object_classrecipientspendingspending_by_awardspending_over_timesubawardstotal_budgetary_resourcestoptier_agenciesBEFORE YOU INSTALL
What to know first
Tools you get
10 tools named in the project's documentation.
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
Start the server by one of: uv run src/server.py (local), uvx --from usaspending-mcp-server@latest usaspending-mcp-server (PyPI), or docker build -t usaspending-mcp-server . then docker run -p 8000:8000 usaspending-mcp-server
Add the mcpServers block to claude_desktop_config.json pointing npx mcp-remote at http://localhost:8000/mcp
MCP Clients
Any MCP client
{
"mcpServers": {
"usaspending": {
"command": "npx",
"args": [
"mcp-remote",
"http://localhost:8000/mcp"
]
}
}
}