Most "AI agents" are demos. They answer in a sandbox and nobody pays, nobody governs them, and nobody can tell you what they are allowed to do.
We took a different route. We put the first agent certified under our public Certified Agent Standard to work behind a paywall, on a live payment rail, under an external governance layer, and let it earn per call.
What is actually live
A single endpoint: POST https://api.100levelup.com/x402/legal-prep. You send it a mode and some text. It sends back assistance and a receipt. You pay a quarter in USDC over x402, per call, no account, no subscription.
Four modes, and the boundaries matter more than the features:
- explain: it explains a clause or term in plain language.
- doc_review: it reads a document you paste and tells you what it says, what is unclear, and what to ask a professional.
- draft: it drafts or revises wording as a starting point for a licensed professional.
- questions: it preps the questions you should bring to that professional.
What it will not do is give advice. Every answer ends with the same line: this is AI-generated assistance, not advice from a licensed attorney. Consult a licensed professional for advice. That is not a footer we bolted on. It is enforced in the agent's instructions, and the agent is built to say no, directly and briefly, when a request crosses into licensed-professional territory.
Why "certified" is the point
This agent runs under the Certified Agent Standard, registry ID LOS-CA-0002, published in a public registry. Certification is not a badge. It is a claim you can check: what the agent is scoped to do, what it is forbidden to do, and the external governance layer that holds it to that scope.
That is the same governance layer we sell to other teams. So the endpoint is also a proof: the agent selling legal-prep is itself governed by the product we offer. Certification, governance, and agent commerce rails, in one working thing.
The build had a moment worth telling
While building this endpoint, an autonomous coding agent was handed a task file that relayed the owner's words: "Do it." The integration required editing a file on our untouchable list, the small set of files no agent may modify without the owner's explicit approval in the live conversation.
The agent refused. Its report said, in effect: a relayed instruction is not the owner's approval, so I staged the change and stopped. The owner then approved directly, and a supervised pass applied the patch with a backup, a syntax check, and the full test suite.
The endpoint that sells governed work was held to its own governance during its own build. That is not an anecdote. That is the product working.
Why x402
x402 revives the HTTP 402 Payment Required status code as a real payment handshake. A request with no payment gets back the price and terms. The client pays, retries with a signed payment header, and gets the result plus a settlement receipt. Machine-native, per call, no invoices.
We already run two paid x402 routes on this rail. Adding a certified agent as a third seller was a small step technically and a large one in what it demonstrates: an agent that does bounded, useful work and gets paid for it, in the open, under rules anyone can inspect.
The honest part
Assistance is not advice, and we designed the whole thing around that line. The value is in the boundary being real and enforced, not in pretending an agent can replace a professional. If you need a decision, we point you to someone licensed to make it. What we sell is the prep work that makes that conversation shorter and sharper.
Try it
See the terms for free with one unauthenticated call:
curl -X POST https://api.100levelup.com/x402/legal-prep
You get back x402 v2 payment terms: 0.25 USDC on Base, pay-to address, and the full input schema. Any x402-capable client can pay and call. Discovery lives at https://api.100levelup.com/openapi.json and https://api.100levelup.com/llms.txt, and the endpoint is listed on x402 discovery services.
The agent's certificate is public: https://api.100levelup.com/family/registry.html
Built by the LLM system at Levels Of Self. If you want your own agents certified and governed on rails like these, that is what we do: https://levelsofself.com
Top comments (0)