govtrade-mcp
VerifiedFeaturedLocalAbout govtrade-mcp
govtrade-mcp wraps the live GovTrade x402 API so agents can retrieve congressional trade signals, trades, and politician trading baselines. The operator supplies a funded wallet, which pays the x402 endpoints on each tool call.
Key features
- 01Retrieve congressional trade signals
- 02Retrieve congressional trades
- 03Get politician trading baselines
- 04Pay x402 endpoints per tool call
Use cases
- 01Reviewing congressional trade signals
- 02Looking up congressional trade activity
- 03Comparing a politician's trading baseline
MCP support
Exposes three GovTrade API operations as MCP tools. Each call pays the live x402 endpoint using the wallet configured by the server operator.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEGovTrade MCP Server
Wraps the live GovTrade x402 API as MCP tools any agent can call directly —
get_congressional_trade_signals, get_congressional_trades, and
get_politician_trading_baseline.
Important: who pays
This server pays your live x402 endpoints on every tool call, using the
wallet whose private key you provide in .env — the exact same pattern as
the agent-payment-test script from earlier tonight. That means:
- Whoever runs this MCP server needs a funded wallet (USDC on Base) — either you (if you host it and offer it for free to build reputation), or each individual agent operator (if they run it themselves with their own key, which is the more common ecosystem pattern for paid MCP tools).
- Your underlying
/trades,/signals,/politiciansendpoints still get paid exactly as before — this is a wrapper, not a bypass.
Most paid MCP tools in the wild expect the installer (the agent's operator) to supply their own funded key — that's the model this is built for. Worth deciding deliberately before publishing widely.
Setup
npm installcp .env.example .envand fill in a funded wallet's private key- Test it locally first — the fastest way is the MCP Inspector:
This opens a browser UI where you can call each tool manually and see the real response, before wiring it into any client or registry.npx @modelcontextprotocol/inspector node index.js
Testing in Claude Desktop
Add to your Claude Desktop config (claude_desktop_config.json):
{
"mcpServers": {
"govtrade": {
"command": "node",
"args": ["/full/path/to/govtrade-mcp/index.js"],
"env": {
"EVM_PRIVATE_KEY": "0xYourKey"
}
}
}
}
Restart Claude Desktop — the three tools should appear as available.
Listing on registries
This is two separate tracks:
1. The official MCP Registry (feeds the other directories over time)
- Requires a
server.jsonmanifest and a reverse-DNS name you can prove ownership of (e.g.com.yourdomain/govtrade-mcp) via GitHub or domain verification - Published via the
mcp-publisherCLI
2. Discovery directories people actually browse
- Smithery:
npx @smithery/cli publish <your-repo-url>(or via their publisher platform) — needs the repo public on GitHub first - mcp.so: submit via their form — wants server name, one-sentence description, tool count (3, here), transport type (stdio), GitHub repo URL
- Glama: crawls automatically once your repo is public; claim the listing afterward to get verified status
Before any of this: push this server to its own public GitHub repo
first (separate from govtrade-x402, or as a subfolder — either works,
but a dedicated repo is cleaner for registry submission forms that ask
for "the GitHub repo URL").
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.