Ground agents in verifiable US trade figures from FRED — exports, imports and the balance since 1992 — so economic answers cite real data.
An agent asked about prices or rates will happily invent a figure. Enable Trade Balance and it calls the tradebalance tool to check the live number before it answers — so the reply rests on real data, not the model's memory.
Give a market assistant a source it can actually cite: every figure it reports carries Trade Balance as its origin, so a human can trace where the number came from.
Because the model pulls trade balance at the moment it responds, the answer reflects the market now — not whatever was true when the model was trained.
Point your agent framework at the VerveContext endpoint. Every enabled source — this one included — becomes a tool the model can call. No per-source plumbing.
{
"mcpServers": {
"vervecontext": {
"url": "https://api.vervecontext.com/v1/mcp",
"headers": { "x-api-key": "vc_live_••••" }
}
}
}Each call returns structured fields your agent can read — and trace back to this source.
yearMonthyearmonthcustomsDutiesupdatedAtHow to ground an agent in it over MCP.