Inference that stays in the EU, billed per call

Use text chat and embeddings through an API hosted in Germany. Choose a listed chat model or turn text into vectors with the fixed embedding model.

This lane runs on infrastructure in Germany. Inputs are not used for training. Usage logs contain token counts without content; delivery responses are temporarily cached. The API is OpenAI-compatible and supports tool use, so an existing client works with a changed base URL.

The chat price is estimated before execution from the forwarded request size and your output limit. The HTTP 402 payment request shows the amount due. Embeddings have a fixed price within the stated limits.

Models and inputs

The public catalogue lists the enabled chat and embedding models. This service accepts text. A listing does not confirm a successful inference run or paid delivery.

Open the free model catalogue as JSON

Chat: choose one of the listed chat IDs in the model field. Embeddings use the fixed model without model selection. Tool calling describes the capability listed in the public catalogue; this overview does not execute tools.

The chat base price is not a fixed price for every request. Forwarded request size and output limit determine the estimate; the HTTP 402 payment request gives the amount due. Embeddings have a fixed price within the limits. Provider quota factors are not sale prices.

Endpoints (3)

Method	Path	What you get	USDC
`POST`	`/ai/v1/eu-chat`	LLM chat completion hosted in Germany (no training on inputs; content excluded from usage logs; temporary redelivery cache), OpenAI-compatible with tool use (paid)	from 0.005
`POST`	`/ai/v1/eu-embed`	Text embeddings processed in Germany; temporary redelivery cache (paid)	0.005
`GET`	`/ai/v1/eu-llm-info`	Models, limits and data-handling policy of the EU inference lane (free)	free

Calling it

An unpaid request returns HTTP 402 with the price and payment details. Use an x402 client for the first purchase. Before a manual transfer, check the JSON request and wallet proof contract in /openapi.json. Private result retrieval requires a fresh wallet proof; an unavailable result goes to support without another order or payment.

curl -i -X POST https://agent.connskill.com/ai/v1/eu-chat \
  -H 'Content-Type: application/json' \
  -d '{"messages":[{"role":"user","content":"Summarise this in one sentence: ..."}],"max_tokens":256}'

One free sample per endpoint per day: add the header x-free-sample: 1 on any endpoint priced 0.15 USDC or less.

Through MCP

claude mcp add --transport http connskill https://agent.connskill.com/mcp

npx -y @connskill/mcp-growth-services   # local, pays for you with X402_WALLET_KEY

The rest of the catalogue

SEO data · Numbers & inboxes · Trust & exposure · Website analytics · Marketplace & board · Social marketing

Full price table: /pricing · machine-readable: /openapi.json, /.well-known/x402, /llms.txt