# Token Harbor

> Token Harbor is an AI gateway: one API key and one balance to call large
> language models from OpenAI/GPT, Anthropic/Claude, Qwen/Alibaba, Z.AI/z.ai, DeepSeek, Xiaomi/MiMo, Google/Gemini, Meta/Muse Spark, Moonshot AI/Kimi, xAI/Grok and Mistral. It is OpenAI- and Anthropic-API
> compatible, pay-as-you-go, with transparent per-token pricing and automatic
> fallback to another provider serving the same model for reliability.

## Key facts

- One key, every model. The OpenAI-compatible base URL is https://tokenharbor.ai/v1 and the Anthropic-compatible base URL is https://tokenharbor.ai.
- Works with the AI coding agents you already use — Claude Code, Codex, Cursor, Cline, opencode, Aider, Continue, Windsurf, Zed, pi and more — via one command (`curl -fsSL https://tokenharbor.ai/connect.sh | sh`) or manual setup.
- Pay-as-you-go. A free account and a standing free tier on select models (their model IDs end in `:free` and never charge your balance). No sign-up credit. No subscription required.

## Start here

- [Home](https://tokenharbor.ai): what Token Harbor is.
- [Models](https://tokenharbor.ai/models): the full catalog with live prices and an intelligence ranking.
- [Pricing](https://tokenharbor.ai/pricing): pay-as-you-go rates and the monthly Pass plans.
- [Providers](https://tokenharbor.ai/providers): every model provider available through one key.
- [Connect an agent](https://tokenharbor.ai/connect): one command to point your AI coding agent at Token Harbor.
- [Docs](https://tokenharbor.ai/docs): quickstart, API reference, billing, and per-agent setup guides.
- [FAQ](https://tokenharbor.ai/faq): what Token Harbor is, how routing works, billing, and privacy.
- [Halyard](https://tokenharbor.ai/halyard): our desktop agent — reads and edits your files, runs commands, one key for every model.
- [AI Blog](https://tokenharbor.ai/blog): model release news, head-to-head comparisons, and pricing analysis.
- [Verify a request](https://tokenharbor.ai/verify): prove which model served any request — no silent model swaps.
- [Status](https://tokenharbor.ai/status): live uptime for the gateway and every dependent service.

## Providers

- [OpenAI API](https://tokenharbor.ai/providers/openai): 15 OpenAI models through one key.
- [Anthropic API](https://tokenharbor.ai/providers/anthropic): 10 Anthropic models through one key.
- [Qwen API](https://tokenharbor.ai/providers/qwen): 8 Qwen models through one key.
- [Z.AI API](https://tokenharbor.ai/providers/z-ai): 6 Z.AI models through one key.
- [DeepSeek API](https://tokenharbor.ai/providers/deepseek): 5 DeepSeek models through one key.
- [Xiaomi API](https://tokenharbor.ai/providers/xiaomi): 4 Xiaomi models through one key.
- [Google API](https://tokenharbor.ai/providers/google): 3 Google models through one key.
- [Meta API](https://tokenharbor.ai/providers/meta): 3 Meta models through one key.
- [Moonshot AI API](https://tokenharbor.ai/providers/moonshot): 3 Moonshot AI models through one key.
- [xAI API](https://tokenharbor.ai/providers/xai): 3 xAI models through one key.
- [Mistral API](https://tokenharbor.ai/providers/mistral): 2 Mistral models through one key.

## Docs

- [Quickstart](https://tokenharbor.ai/docs/getting-started/quickstart): first API call in minutes.
- [Chat APIs](https://tokenharbor.ai/docs/api/curl): OpenAI /v1/chat/completions and Anthropic /v1/messages examples.
- [Manual agent setup](https://tokenharbor.ai/docs/integrations): configure any agent by hand (Base URL + key).
- [Connect CLI](https://tokenharbor.ai/docs/connect): automatic one-command setup.