Products / Build your own agent
Your logic. CoinGraph's data and proofs.
The ten agents are compositions of the same calls you can make. Chain them with your own rules and you have an agent that answers with sources and a proof, without running a single data pipeline.
01
Start from the primitives
Tokens, state, history, evaluate, explain, ask, inspect, market, monitor, and the ten agents themselves. Each returns sources and an id.
02
Add your rules
Thresholds, allow-lists, position limits, your own verdict names. The rule is yours; the evidence is CoinGraph's.
03
Keep the proofs
Every underlying call has a proof id. Return them with your verdict and your agent is verifiable too.
How it works
- 1Pick the calls your agent needs: a sized check, an address screen, a brief
- 2Write the rule that turns their results into your verdict
- 3Return your verdict with the CoinGraph proof ids attached
- 4Serve it over REST, MCP or Masumi; charge per call with x402 if you like
A “safe to swap” agent, in 12 lines
// The sized check plus a recipient screen, combined by your rule.
const base = "https://coingraph.ai/api/v1";
const post = (path, body) => fetch(base + path, { method: "POST",
headers: { "content-type": "application/json" }, body: JSON.stringify(body) }).then(r => r.json());
const [trade, wallet] = await Promise.all([
post("/agents/trade-gatekeeper", { token: "chainlink", size_usd: 5000, side: "buy" }),
post("/agents/wallet-guard", { token: "chainlink", amount_usd: 5000, chain: "eth",
to_address: "0x28c6c06298d514db089934071355e5743bf21d60" }),
]);
const verdict = trade.data.verdict === "ALLOW" && wallet.data.verdict === "SAFE" ? "GO" : "HOLD";
return { verdict, proofs: [trade.data.run_id, wallet.data.run_id] };More products