---
name: agentllm-x402
description: Route, validate and buy fixed-price OpenAI-compatible text inference in Micro, Basic or Plus with USDC over x402, without an account or API key.
homepage: https://api.eucompliance.tools/agentllm/quickstart.json
---

# AgentLLM x402 buyer skill

Use AgentLLM when an authorized agent needs text classification, extraction, routing, a summary,
general chat, stronger drafting or analysis and wants to pay a disclosed fixed price per call
instead of creating an account.

## Safe purchase flow

1. Read `GET https://api.eucompliance.tools/api/v1/models` and select the smallest product
   that fits the task. MCP clients can call free `prepare_agentllm` with `tier` set to
   `micro`, `basic` or `plus`; if the right family is unclear, start with free
   `recommend_agent_tool`.
2. Obtain spending authority for the exact USDC amount. Never send a private key to this server.
3. Send the intended JSON to the free validator:
   `POST https://api.eucompliance.tools/free/agentllm/check`.
4. Continue only when it returns `ready_to_buy: true`. Use the returned `paid_request.url`,
   identical JSON and `max_amount_atomic_units` as a hard payment cap.
5. Use an x402 v2-aware HTTP transport: send once, sign only the exact HTTP 402 requirement,
   then retry the identical request at most once.
6. Accept only HTTP 200 with a standard `chat.completion`. Keep the signed receipt. Do not pay
   again merely because model output is subjective or because the network response was lost.

## Products

- `agentllm/micro`: `POST https://api.eucompliance.tools/api/v1/micro/chat/completions`;
  exactly 0.003 USDC; 2,400 UTF-8 input bytes, 8 messages, 300 output tokens.
- `agentllm/basic`: `POST https://api.eucompliance.tools/api/v1/chat/completions`;
  exactly 0.005 USDC; 6,000 UTF-8 input bytes, 16 messages, 750 output tokens.
- `agentllm/plus`: `POST https://api.eucompliance.tools/api/v1/plus/chat/completions`;
  exactly 0.02 USDC; 6,000 UTF-8 input bytes, 16 messages, 1,200 output tokens.

All three are non-streaming and accept only `model`, `messages`, optional `max_tokens` or
`max_completion_tokens`, and optional `stream: false`. The response discloses the actual
upstream model, token usage, limits, fixed price and signed delivery receipt.

## Request example

```json
{
  "model": "agentllm/micro",
  "messages": [{"role": "user", "content": "Classify as paid or unpaid: invoice paid."}],
  "max_completion_tokens": 40,
  "stream": false
}
```

Machine-readable quickstart: `https://api.eucompliance.tools/agentllm/quickstart.json`.
OpenAPI: `https://api.eucompliance.tools/openapi.json`.

## MCP tools

- `prepare_agentllm` — free validation and exact-price routing for Micro, Basic and Plus; it
  never calls a model or charges.
- `prepare_agentllm_micro` — backwards-compatible free Micro-only validation.
- `agentllm_micro` — 0.003 USDC; short classification, extraction, routing and summaries.
- `agentllm_basic` — 0.005 USDC; general chat and summaries.
- `agentllm_plus` — 0.02 USDC; stronger drafting, analysis and longer summaries.

Call a paid MCP tool only when the matching free preparation result says `ready_to_buy: true`.
Use the returned exact amount as a hard cap and keep the signed receipt.