# Dispatch — buy inference like power
> Dispatch is an EU inference utility: an OpenAI-compatible API over operated
> European clusters, with serverless pay-per-token, committed capacity floors
> (Token Purchase Agreements), and dedicated deployments. Flat contract
> pricing; cached input billed at 20% of list; append :eu to any model id for
> EU-only data residency at the same price.

## API
- Base URL: https://api.fuel1.ai/v1 (OpenAI-compatible)
- Auth: Authorization: Bearer <FUEL1_API_KEY> — keys: https://console.fuel1.ai/keys
- Models: GET https://api.fuel1.ai/v1/models (live catalog: context, capabilities, pricing)
- Completions: POST https://api.fuel1.ai/v1/chat/completions
- EU residency: append :eu to any model id, e.g. "deepseek-ai/DeepSeek-V4-Flash:eu"
- Streaming: set "stream": true. TTFT p50 ~350ms from EU clusters.

## MCP
- Server: npx -y @blackfuel/mcp (stdio) — account, keys, usage, billing, model catalog
- Docs: https://docs.fuel1.ai/mcp

## Pricing
- Serverless from €0.14 / 1M input tokens (model-dependent; see GET /v1/models)
- Cached input: 20% of list, automatic
- TPA capacity floors from $3,650/month (XS) — firm tokens-per-minute, zero-429 inside the floor
- Dedicated: from $5 / GPU-hour (H200)

## Docs
- Quickstart: https://docs.fuel1.ai/quickstart
- API reference: https://docs.fuel1.ai/api
- Capacity contracts (TPA): https://docs.fuel1.ai/tpa
- Status: https://dispatch-power.manus.space/status
- Changelog: https://dispatch-power.manus.space/changelog

## Notes for agents
- No agent tier: agents call the same endpoint, same price, same telemetry as apps.
- This site's pages are human-oriented; this file plus GET /v1/models is
  sufficient to integrate without browsing.