Read-only diagnostics for AI agents and Web3 payments. Every command answers one question with PASS / FAIL / UNKNOWN — and documents what it does not cover.
A real, running MCP server for cryptographically verifiable receipts and agent payments: after an x402 settlement, it anchors a SHA-256 digest of delivered content on Solana so anyone can verify delivery offline, forever. (Formerly known as AgentToll.)
Not affiliated with receiptrail.ai — a separate funded SaaS for bank-card chargeback management. ReceiptRail is an independent open-source x402 delivery-receipt service on Solana. We do not compete with or critique that product; the two are different entities.
Identifiers: ReceiptRail · npm receiptrail-mcp (v0.3.0) · GitHub xka0085-byte/agenttoll · official MCP Registry io.github.xka0085-byte/receiptrail.
Is your paid endpoint advertising itself the way agents expect? One inspect command parses the HTTP 402 payment document (v1/v2 accepts[]), checks the /.well-known/mcp and /.well-known/x402 discovery manifests, and returns a response SHA-256. Read-only, not an MCP server.
Not affiliated with mcpdoctor.dev — a separate commercial product that audits MCP servers. This is an independent, read-only 402/x402 endpoint inspector CLI.
What does an x402 endpoint demand before serving? Parses HTTP 402 payment documents: scheme, network (CAIP-2), asset, amount, payTo — plus response latency and SHA-256 evidence. Never pays.
Does your MCP server advertise OAuth metadata an agent can actually discover? 401 challenge, WWW-Authenticate resource_metadata, Protected Resource Metadata, authorization server metadata, PKCE S256.
Solana transaction evidence from one public RPC: slot, block time, execution error, parsed instructions, fee, response hash. If the RPC cannot see a transaction it says UNKNOWN — never "does not exist".
Normalizes operator-supplied cross-chain message status JSON (source/destination/delivered/failed) into a conservative evidence report. Missing fields stay UNKNOWN. Protocol APIs are the roadmap.