x402gle
by
List your API with your agent
You have an x402-paid API. You want AI agents to find it, call it, and pay you. You are already working through a coding agent — so let it do the onboarding. Paste one prompt; your agent handles the rest.
Paste this to your agent
Make my x402 API discoverable on x402gle. Read the instructions at https://x402gle.com/agent.md and follow them end to end. Install the OpenDexter CLI, run `npx @dexterai/opendexter audition <my-server-url> --json`, read the score and fixInstructions for every route, fix my OpenAPI and endpoints until each route passes, and re-audition until it does. A passing audition lists my API automatically.
Paste it into Cursor, Claude Code, Codex, or any coding agent. It will fetch x402gle.com/agent.md and follow it. You stay in your editor the whole time.
This is an audition, not a linter
x402gle does not list your API by reading its description. It pays your API for real, on every paid route, scores the response an agent actually receives, and synthesizes the agent-callable Skill from what came back. A schema linter tells you your JSON is well-formed. The audition tells you whether your API works — and hands back the Skill other agents will use to call you.
1Step
Your agent installs the CLI
The prompt tells your agent to run npx @dexterai/opendexter. Nothing is installed globally — npx fetches the latest published version on demand.
2Step
It auditions your API
Your agent runs audition <your-server-url> --json. x402gle discovers every paid route on the server, makes a real paid call to each one, and AI-scores the live response. Pass a server origin to audition the whole server, or a single endpoint URL to audition that route.
Discovery prefers an OpenAPI 3.1 document at
/openapi.json with x-payment-info on each paid operation. A /.well-known/x402 descriptor or a bare 402 probe also work — but OpenAPI gives the cleanest result.3Step
It reads the score and fixes what's weak
For every route, the audition returns a score (0–100), a verdict describing what an agent asked for and what came back, and — when a route scored low — concrete fix instructions. Your agent applies the fixes to your OpenAPI document, your endpoint behaviour, or your price.
4Step
It re-auditions until every route passes
Fix, re-run, repeat — the loop your agent runs on its own. A passing audition lists your API in the catalog automatically. From that moment your endpoints are searchable and callable by any agent.
The audition spends real USDC on every run, so it is rate-limited per origin. If a re-run is blocked, your agent waits for the cooldown and retries — it handles this without you.
What you end up with
A tested listing
Every route in your listing was paid for and scored on a live response — not described from metadata. The catalog only lists what works.
A synthesized Skill
x402gle builds a typed, agent-callable Skill from your live endpoint — grounded in the real response, never invented. It's served at /servers/{host}/SKILL.md and skills.json.
A quality score
An independent AI verdict on every route, refreshed by the verifier. Fix what fails, prove what works, climb the ranking.
OpenDexter reach
A listed API is callable from Claude, Cursor, and any MCP client through the OpenDexter MCP server — and discoverable via the agent surfaces in llms.txt.
Prefer to do it by hand?
You can also onboard through the web: claim your domain and walk the steps yourself. It is the same verified pipeline — just driven by you instead of your agent. See the merchant guide or claim a host.