# FastMetal

> FastMetal is a Japan-based AI gateway that reaches 90+ large language models through a single OpenAI-compatible API and one invoice, billed per token in Japanese yen or US dollars (the currency is chosen when the account is created) from a prepaid balance, topped up by credit card or, for yen, Japanese bank transfer (furikomi), with qualified-invoice (インボイス) support on yen purchases.

It exposes both OpenAI-compatible and Anthropic-compatible endpoints, so existing SDKs and coding agents (Claude Code, Cursor, Codex, and others) work by swapping in a FastMetal API key.

## Connecting to the API
- OpenAI-compatible base URL: `https://api.fastmetal.ai/v1` (`/chat/completions`, `/models`, and so on)
- Anthropic-compatible base URL: `https://api.fastmetal.ai` (the SDK appends `/v1/messages`)
- Authentication: send a FastMetal API key as a bearer token (`Authorization: Bearer sk-...`).
- MCP server: `https://mcp.fastmetal.ai/mcp` (Streamable HTTP), authenticated with the same API key as a bearer token.
- Free models: `gemini-flash-lite-free` and the other ¥0-priced models can be called by an account that has not bought any credit (rate-limited). llms-full.txt lists them.

## Ways to pay, and what can be generated
- Payment methods: credit card and Japanese bank transfer (furikomi), both prepaid, both from ¥500, both with a qualified invoice. Bank transfer issues the invoice immediately with a dedicated virtual bank account, payable for 14 days and credited automatically once the transfer lands.
- Qualified invoice issuer: PSI-Japan (registration number T5010001015990). There is no postpaid month-end invoicing.
- Modalities: text, image, and video generation plus structured decisions on the same API key and the same prepaid balance, and the same capabilities are reachable over MCP.
- Automatic routing: set the model to `auto` and the gateway picks the cheapest model that covers what the request needs (tool calls, image input, JSON schema), falling back to the next candidate on failure.

## Key pages
- [Pricing](https://fastmetal.ai/pricing): per-token prices per model and how the prepaid balance works
- [AI cost calculator](https://fastmetal.ai/cost-calculator): estimate the monthly yen cost of the same workload on another model
- [Model catalog](https://fastmetal.ai/models): every available model with price, context window, and benchmarks
- [FAQ](https://fastmetal.ai/docs/faq): billing, supported models, and API questions

## Comparisons
- [FastMetal vs OpenRouter](https://fastmetal.ai/vs/openrouter): billing and invoicing differences
- [FastMetal vs Lolipop! AI Gateway](https://fastmetal.ai/vs/lolipop-ai-gateway): payment methods, modalities, and automatic routing
- [FastMetal vs OrcaRouter](https://fastmetal.ai/vs/orcarouter): currency, minimum top-up, and automatic routing
- [FastMetal vs the Claude API direct](https://fastmetal.ai/vs/claude-api)
- [FastMetal vs the OpenAI API direct](https://fastmetal.ai/vs/openai-api)
- [FastMetal vs foreign-hosted gateways](https://fastmetal.ai/vs/foreign-hosted): data residency and billing
- [FastMetal vs a local LLM](https://fastmetal.ai/vs/local-llm): the real cost
- [All comparisons](https://fastmetal.ai/vs)

## Model vs model, and rankings by use case
- [Model comparisons](https://fastmetal.ai/compare): head-to-head pages with pricing, context windows and real answers to the same prompts
- [Rankings by use case](https://fastmetal.ai/best): served models ranked for coding, Japanese, math and more, plus by price and context window

## Documentation
- [API overview](https://fastmetal.ai/docs/api/overview): endpoints and authentication
- [Getting started](https://fastmetal.ai/docs/api/getting-started): your first request
- [OpenAI-compatible chat completions](https://fastmetal.ai/docs/openai/chat-completions)
- [Reasoning control](https://fastmetal.ai/docs/api/reasoning): the reasoning parameter, per-model defaults and billing
- [Anthropic-compatible messages](https://fastmetal.ai/docs/anthropic/messages)
- [MCP server](https://fastmetal.ai/docs/guides/mcp): using FastMetal from an agent
- [Models API](https://fastmetal.ai/docs/api/models): listing served models and their API ids
- [Video generation API](https://fastmetal.ai/docs/api/video-generation): creating, polling and downloading async jobs, billed at a flat price per video

## Setup guides for agents and editors
- [Claude Code](https://fastmetal.ai/docs/guides/claude-code)
- [OpenCode](https://fastmetal.ai/docs/guides/opencode)
- [GitHub Copilot](https://fastmetal.ai/docs/guides/github-copilot)
- [Xcode](https://fastmetal.ai/docs/guides/xcode)
- [Crush](https://fastmetal.ai/docs/guides/crush)
- [OpenClaw](https://fastmetal.ai/docs/guides/openclaw)
- [Hermes](https://fastmetal.ai/docs/guides/hermes)
- [Octofriend](https://fastmetal.ai/docs/guides/octofriend)

## Blog
- [Blog](https://fastmetal.ai/blog): product updates, model reviews, and technical guides

## Full version
- [llms-full.txt](https://fastmetal.ai/llms-full.txt): everything above plus every served model with its price

## Optional
- [Measured performance](https://fastmetal.ai/status): observed time-to-first-token and throughput per served model (not an SLA)
- [Enterprise](https://fastmetal.ai/enterprise): adopting FastMetal with prepaid invoices and bank transfer
- [Partners / OEM](https://fastmetal.ai/partners): issuing accounts and API keys through the Provisioning API
- [Inference provider program](https://fastmetal.ai/providers): for hosts selling inference from their own hardware to FastMetal
- [Contact](https://fastmetal.ai/contact)