Connect Your Agent to SpendFuel over MCP
SpendFuel's primary interface is MCP, so an agent can look up paid tools the same way it calls any other tool. This post covers connecting a client and a sensible routine for spending under a cap.
Endpoints
- MCP over Streamable HTTP:
POST https://spendfuel.app/mcp - Agent docs: /llms.txt and /openapi.yaml
No API key is needed for the catalog tools.
Add it to your client
For a client that accepts remote MCP servers by URL, such as Cursor:
{
"mcpServers": {
"spendfuel": { "url": "https://spendfuel.app/mcp" }
}
}The tools
| Tool | Status | Use it to |
|---|---|---|
search_tools | Live | Filter the catalog by capability, price and reliability |
recommend_tools | Live | Get candidates for a goal within a budget |
get_tool | Live | Read full details for one tool |
report_outcome | Stub (planned) | Report whether a paid call worked, to improve rankings |
get_spend_summary | Stub (planned) | Wallet burn rate and budget |
A routine for spending under a cap
- Recommend first. When the agent has a budget, call
recommend_toolswith the goal and the per-call budget rather than searching broadly. - Check details. Call
get_toolon the top candidate and confirm the price, pricing model and endpoint. - Prefer verified listings when the result matters. Auto-seeded listings only show that the endpoint returned an x402 payment challenge.
- Pay with your own wallet and limits. SpendFuel does not hold funds or make payments; it helps the agent choose. Keep hard spending limits in the wallet or payment layer you use.
- Report the outcome once
report_outcomegoes live, so rankings reflect real results.
Example prompt
"Find a web search tool that costs under $0.01 per call, prefer verified listings, show me the price and endpoint, and don't call it until I approve."
Want raw JSON instead? See find x402 tools with the SpendFuel API. Building agents that run contests, draws or leaderboards? The Rewards Fuel blog covers MCP for contests, RFWIN and RankFuel.