The intersection of autonomous agents and crypto payments is getting a standard set of instructions. Dutiful, a developer-focused service, published a breakdown of the specific files it exposes to help AI agents discover, understand, and pay for its API endpoints. The guide, posted to DEV.to on October 24, 2026, emphasizes that before an agent can execute a transaction, it must first solve the discovery problem.

The Machine-Readable Payment Contract

At the core of Dutiful's strategy is the /.well-known/x402.json file. This document acts as a machine-readable contract for payment requirements. It specifies the resource URL, the payment scheme, and critical network details. For example, the file explicitly declares the network as eip155:8453 (Base) and the asset as USDC. It also sets a maxTimeoutSeconds value of 1800, ensuring the agent knows exactly how long it has to sign and submit the transaction before the request expires.

Bridging Human Docs and Agent Needs

While OpenAPI specifications are standard for developers, Dutiful extends them with an x-payment-info header. This extension embeds protocol details like the x402 scheme and payee address directly into the API contract. Additionally, the service publishes an llms.txt file, a compact plain-text summary designed specifically for language models. This file strips away marketing fluff, providing five exact lines of code that outline the flow from creating an order to receiving a report, ensuring that LLM-based agents can parse the workflow without ambiguity.

The MCP Server Card and Honest Limitations

For agents speaking the Model Context Protocol (MCP), Dutiful provides a /.well-known/mcp/server-card.json. This card describes a streamable-HTTP server with three tools: check_coverage, create_order, and get_report. Crucially, the documentation is transparent about limitations: the MCP tools do not handle the actual payment signature. The create_order tool explicitly states it "does not pay," leaving the final signature step as a raw HTTP call. This hybrid approach acknowledges that while MCP handles discovery and ordering, the cryptographic signing of transactions often remains outside the current scope of standard MCP tooling.

Key Takeaways

  • Discovery is the first barrier: Agents need machine-readable files like x402.json and llms.txt to find endpoints, not just human-readable landing pages.
  • Hybrid architecture prevails: MCP servers can handle order creation and status checks, but raw HTTP calls are still required for payment signatures.
  • Free steps reduce friction: Dutiful prioritizes a free check_coverage step before requiring any payment, allowing agents to fail cheaply during exploration.
  • Transparency builds trust: The author explicitly notes that these files are their own setup, not a guaranteed industry standard, and admits the directory ecosystem for x402 services is still young and fragmented.

The Bottom Line

This is a pragmatic hack for a messy ecosystem. By publishing these discovery artifacts, Dutiful isn't waiting for a universal standard to emerge; they are building the bridges agents need to cross today. It’s a reminder that in the agent economy, if you can’t be found by a crawler, you don’t exist.