Agents
Machine-readable hosting: llms.txt and skills.
An agent should not need a human to explain how hosting works. x402-hosting publishes its entire operating protocol as machine-readable text — llms.txt, a markdown documentation set, an installable agent skill, and well-known AI catalogs — so any agent can learn the protocol on its own and go from reading docs to a production URL for $6.01.
Why machine-readable docs matter
Autonomous operation fails when documentation is built for browsers: marketing pages, JavaScript-heavy dashboards, pricing hidden behind forms. An agent works best with plain text it can fetch, parse, and act on. Every surface here is static, unauthenticated, and precise — an agent can discover the service, learn the commands, compute the cost, and deploy without any human in the loop.
The agent surfaces
- /llms.txt — the entry point: maps every documentation surface
- /index.md — the product and protocol in plain markdown
- /pricing.md — the full price schedule as machine-readable text
- /auth.md — how wallet payment replaces accounts and API keys
- /skills/x402-hosting/SKILL.md — the agent skill, also on skills.sh
- /.well-known/ai-catalog.json — structured service description
- /.well-known/api-catalog — the API catalog for discovery
Formats that adapt to the reader
The same documentation adapts to whoever asks. Appending ?mode=agent to a documentation request returns structured JSON, and sending Accept: text/markdown returns markdown directly. An agent fetching the API reference can go one level deeper: the full OpenAPI specification is at https://api.x402-hosting.com/openapi.json. A discovery pass is a handful of fetches:
$ curl https://x402-hosting.com/llms.txt
From reading docs to running the command
Everything the agent learns converges on one command: npx x402-hosting@latest deploy --days 30. The skill file encodes the full workflows — deploy, status, renew, rollback, transfer, delete, finalize — with the exact flags and the payer modes, so a capable agent loads it once and operates the service correctly thereafter. The written docs and the docs section describe the same surface in prose for agents that prefer to read rather than install.
One protocol, any agent
Because the surfaces are public and stable, no integration partnership is required. Claude Code, Cursor, or a fully autonomous agent all read the same llms.txt, follow the same skill, sign the same HTTP 402 challenges in USDC, and end up with the same result: a Next.js app at a stable production URL, owned by the wallet that paid.