Developers

API keys aren't open yet. This page shows how they'll work; tell us if you'd like early access.

What a key reaches

Read-only data, no AI on our side, 5,000 calls a key a day (the count resets at 00:00 UTC; every reply carries X-Quota-Limit, X-Quota-Remaining and X-Quota-Reset in seconds, and Retry-After once the day's calls are used). A key never reaches your pursuit desk, saved searches or account.

Ask with a plan

The site turns questions into a plan (a small JSON: what to measure, how to group, which filters). Send a plan and get the rows, the SQL that ran and the caveats. The plan's fields are in the catalog the MCP connector's get_catalog tool returns.

curl -s https://api.spendquery.com/plan \
  -H "Authorization: Bearer sqk_your_key" \
  -H "content-type: application/json" \
  -d '{"intent": "aggregate", "metric": "obligations", "group_by": ["recipient"],
 "fiscal_years": [2026], "filters": [{"dimension": "agency", "values": ["Department of Energy"]}],
 "limit": 10}'

Lists

curl -s "https://api.spendquery.com/opportunities?naics=541512&limit=20" -H "X-API-Key: sqk_your_key"
curl -s "https://api.spendquery.com/recompetes?months=12&min_value=1000000" -H "X-API-Key: sqk_your_key"
curl -s "https://api.spendquery.com/sbir?agency=Department+of+Energy&phase=Phase+II" -H "X-API-Key: sqk_your_key"
curl -s "https://api.spendquery.com/rules?area=far&kind=proposed" -H "X-API-Key: sqk_your_key"

MCP connector for AI assistants

Add SpendQuery as a remote MCP server in Claude, ChatGPT or any MCP client: URL https://api.spendquery.com/mcp. No sign-in or key is needed (600 calls a day per network address); with an API key as a bearer token you get your key's own allowance. The assistant plans the question itself; SpendQuery runs it read-only, within a time limit, and every number comes with the SQL or the record behind it.

Tables and charts in the chat

In hosts that support MCP Apps (Claude, ChatGPT, VS Code), four tools answer with an interactive view instead of raw data: open notices as a table with due dates and set-asides (search_notices), a company's federal profile with its agencies and years (company_profile), contracts coming up for re-bid with the incumbent (recompetes) and spending answers as bar charts (run_plan). Links in the views open SpendQuery's pages; the host may ask you once to allow the view. Other clients get the same data as before.

Tools

Prompts

Four prompts (slash commands in Claude Desktop) chain the tools the way SpendQuery's own workflows do: bid_no_bid (a go / no-go memo on one SAM.gov notice), qualified_work (open federal work a company can win, ranked, with what was left out and why), market_research(a draft FAR Part 10 market research report) and sources_sought_reply (a short reply to a sources sought or RFI, past work left as placeholders). Each asks the assistant to take every number from a tool result.

Try asking

What's new

Fair use

The data comes from public sources (USAspending.gov, SAM.gov, SBA, the Federal Register, state and local portals); cite them and SpendQuery. Don't share keys; revoke a key you no longer use. The terms apply. Where the numbers come from.