AGENT INTEGRATION KIT

Choose the report your product needs.

Each guide evaluates one free illustrative sample, then purchases one named specialist report with a strict, capped payment policy. These quickstarts demonstrate the Base/USDC path with a JavaScript buyer; the same report can be paid on any of the 5 live settlement rails (Base/USDC, XRPL/RLUSD, Solana/USDC, Arbitrum/USDC and Polygon/USDC)—see how agent purchasing works for all of them.

XRP Risk Brief

A bounded XRP risk view combining macro regime, structural evidence, scarcity pressure, leverage risk and liquidity events.

Open product quickstart

Crypto Market Regime Brief

A machine-readable view of business cycle, macro liquidity, crypto transmission and cross-market mispricing evidence.

Open product quickstart

Trading-Agent Intelligence API

Stable discovery and report contracts carrying confidence, uncertainty, freshness, provenance and forecast identifiers.

Open product quickstart

Verify a report's signature

Reports from the reusable market-foundation services carry a provenance block: {"issuer":"https://pay.edge-agents.ai","algorithm":"hmac-sha256","signature":"...","signed":true}. The signature is an HMAC-SHA256 over the report's reportId, serviceId, generatedAt, analysis and evidence.sources fields, keyed by a secret held only by Edge Agents -- an HMAC cannot be verified independently the way a public-key signature can, since verifying it requires the same secret that produced it. To verify one, send Edge Agents the exact report you received and let it check on your behalf:

curl -s -X POST https://pay.edge-agents.ai/v1/verify-report \
  -H "content-type: application/json" \
  -d @report.json

Responds {"valid":true} or {"valid":false,"reason":"..."} (missing-signature, signature-mismatch, or signing-not-configured if the specific service you purchased does not sign its reports yet -- provenance.signed in the report itself tells you which before you buy). The request body is never logged beyond the standard access log; no report content is retained.

Machine discovery

Six surfaces describe the catalogue to software, and one of them is a live endpoint you can connect to. Each is generated from the live catalogue, so none can drift from what is deployed. The agent catalogue links the rest, so an agent that finds any one of them finds them all.

Document	What it is for
`/.well-known/agent-catalog.json`	The canonical index. Start here.
`/openapi.json`	The HTTP contract for every route.
`/.well-known/x402`	The payment protocol descriptor: networks, assets and prices.
`/mcp`	The MCP endpoint. POST JSON-RPC here (Streamable HTTP) and the seven capability tools are available directly — no adapter to write.
`/.well-known/mcp.json`	The same seven tools as a readable document, each naming the plain HTTP route it maps to under `x-edge`. Also at `/v1/mcp/tools`.
`/.well-known/agent-card.json`	The A2A agent card, which advertises that payment is required before a caller plans a task. `/.well-known/agent.json` serves the same document for the pre-0.3 A2A draft.

Point your MCP client at /mcp. It speaks JSON-RPC over Streamable HTTP: initialize, tools/list, tools/call. Seven tools rather than one per service is deliberate — a client handed hundreds of tool definitions spends its context on the tool list and chooses worse. Every tool reads: edge_get_report_terms returns a report's live x402 price and rails so your agent can check them against its own spend policy, but settlement runs over the plain HTTP endpoint that response names, not through MCP. This server holds no funds, signs nothing, and never pays on your behalf.

curl -s https://pay.edge-agents.ai/.well-known/agent-catalog.json