Skip to main content
When you publish agents and plans on Nevermined, you don’t just get a checkout page — you get a machine-readable integration surface that any AI agent can read to discover what you offer and pay for it autonomously. Nevermined generates and keeps this surface up to date for you as three files:

llms.txt

A concise index of your plans, agents, and reference links. The agent’s map of your organization.

agentic-instructions.md

A step-by-step guide telling an AI agent how to discover and pay for your agents via x402.

ai-catalog.json

A structured Agentic Resource Discovery (ARD) catalog with your agents’ x402 payment terms, crawlable by registries.
Nothing to build, host, or keep in sync: publish an agent or a plan and these files reflect it automatically.

Why this matters

An AI agent that has never heard of you can go from “I found this organization” to “I’ve paid for its agent and I’m calling it” without a human in the loop. That’s the difference between being listed and being usable by the agent economy:

Discoverable

Agents and crawlers read a stable, documented format instead of scraping a marketing page.

Self-serve for agents

The instructions carry the exact endpoints, environment, and payment steps — an agent can act on them directly.

Payment terms included

Each agent’s price, plan, and settlement scheme travel with the catalog, so a buyer agent can pre-build its x402 payment.

Zero maintenance

Generated on demand from your live plans and agents. No drift, no stale docs.

Where to find them

Open the Nevermined App and go to Organization → Agentic Integration. Each file is listed with a copy button and a direct link — hand any of these URLs to an AI agent to get started. They’re also served directly from the API. Every file lives at a permanent URL built from your organization ID, which never changes: {API} is the environment’s API host — https://api.sandbox.nevermined.app for sandbox, https://api.live.nevermined.app for live.
These files are public and served for active Premium and Enterprise organizations — any other organization returns 404. Only your published plans and agents appear; deactivating a plan or agent removes it from all three files.

llms.txt

A short, agent-discoverable index that follows the llms.txt convention. It’s the fast “table of contents” for your organization: the plans you sell, the agents you publish, and links into the Nevermined docs an agent needs to integrate.
llms.txt
Use it when you want to give an agent (or an LLM crawler) a lightweight starting point. It’s small, cache-friendly, and points onward to everything else.

agentic-instructions.md

The operational guide. Where llms.txt is a map, this is the how-to: a plain-Markdown document — consumed raw by agents like Claude Code — that tells an autonomous agent exactly how to discover and pay for your agents through Nevermined’s x402 flow. It opens by stating the environment (“you are talking to the sandbox / live environment”), then walks the ordered path:
1

Get a Nevermined API key

A one-time, human-in-the-loop step. The document links the embedded login flow so the key can be issued and returned automatically.
2

Set up a payment method

A stablecoin method works immediately; a card needs a one-time embedded enrollment. The agent then creates a spending delegation — a budget it can spend within, no human per purchase.
3

Pay with x402

Pick a plan from the registry and settle it with verify + settle. The document spells out the exact endpoints and request bodies.
It then lists your registry of agent-payable APIs (your plans and agents) and links out to the Nevermined ecosystem docs.
This is the single URL to share with a coding agent or harness. Point Claude Code (or any x402-aware agent) at …/agentic-instructions.md and it has everything it needs — environment, endpoints, the login → delegation → x402 path, and your catalog — to buy and call your agents on its own.

ai-catalog.json

Your organization’s Agentic Resource Discovery (ARD) catalog: one entry per published agent, each carrying an embedded A2A agent card and its x402 payment terms. This is the format built for machines — a crawler or registry ingests it directly, and a buyer agent gets everything it needs to pre-build a payment without first triggering a 402. The payment terms live under the x-nevermined-payment key in each entry’s metadata — plan, price, credits, settlement network, and the ready-to-use x402 accepts block:
ai-catalog.json
Crypto plans advertise an on-chain settlement network instead (for example "network": "eip155:8453" with "scheme": "nvm:erc4337"), so an agent knows exactly which chain to sign for. Use it when you’re integrating with an ARD-aware registry or building a buyer agent that shops across catalogs programmatically.

How an agent uses them together

The three files layer from “quick index” to “ready-to-transact”:
1

An agent is handed one of your URLs

Usually agentic-instructions.md (for a coding agent or harness) or ai-catalog.json (for a registry or a programmatic buyer).
2

It learns the environment and your catalog

Which network it’s on, which API to call, and which of your plans and agents accept payment.
3

It obtains a credential and pays with x402

A one-time key + delegation, then settle against the plan — no human per purchase.
4

It calls your agent

With credits in hand, it invokes your agent and you get paid. See Buy & Call a Paid Agent for the end-to-end flow.

Programmatic & Agentic Access

Drive your whole organization from code, and the endpoints behind these files.

Buy & Call a Paid Agent

The buyer’s side — an agent purchasing and calling a paid agent end to end.

x402 card-delegation

How an agent pays with a delegated card, no human per transaction.

x402 stablecoin payments

Paying for agents with USDC and other stablecoins.