DaoXEDaoXE is OpenAI-compatible with a single base URL (https://daoxe.com/v1) and native Anthropic Messages for Claude Code, so most tools need only a base-URL-and-key change. This hub links every integration guide, every model-family guide, and the honest gateway comparisons. Not sure where to start? Use the decision table — and always detect model swapping so a cheaper endpoint can't quietly downgrade you.
Editors & CLI agents#
One page per tool — real, tool-specific setup and caveats, not a copy-paste template:
- Cursor — Cursor via Override OpenAI Base URL + custom models.
- Cline — Cline via the OpenAI Compatible provider.
- Roo Code — Roo Code via OpenAI Compatible, per-mode models.
- Continue — Continue via config.yaml apiBase.
- Aider — Aider via OPENAI_API_BASE + openai/ prefix.
- Zed — Zed via language_models.openai_compatible.
- opencode — opencode via a custom provider in opencode.json.
- Cherry Studio — Cherry Studio via an OpenAI-type provider.
- Open WebUI — Open WebUI via an OpenAI API connection.
By model family#
Reach a specific model family through your one key — with the real per-model quirks:
- Claude — Claude via OpenAI chat + native Anthropic Messages.
- GPT — GPT via chat + Responses; drop-in OpenAI SDK.
- Gemini — Gemini via /v1/chat/completions; no Google key needed.
- DeepSeek — DeepSeek via /v1/chat/completions; mind reasoning_content.
- Kimi (Moonshot) — Kimi via /v1/chat/completions; disable thinking for plain output.
- Qwen — Qwen via /v1/chat/completions; mind enable_thinking.
- Doubao — Doubao via /v1/chat/completions; catalog ID, no ep- setup.
Compare & decide#
Honest, verifiable head-to-heads and cost guides:
- DaoXE vs OpenRouter — OpenRouter — broad hosted multi-provider router.
- DaoXE vs LiteLLM — LiteLLM — self-hosted OpenAI-compatible proxy.
- DaoXE vs One API / New API — One API / New API — self-hosted relay & key-management panels.
- cost vs official APIs
- OpenRouter alternative
- cheapest APIs guide
- gateway benchmark guide
- detect model swapping
Which path should I take?#
Pick the shortest path to what you actually want to do:
| You want… | Best DaoXE path | Guide |
|---|---|---|
| An autonomous coding agent in VS Code | Cline or Roo Code via the OpenAI Compatible provider | Open guide |
| Cursor running your own models | Override OpenAI Base URL + custom model IDs | Open guide |
| Native Claude Code | Anthropic Messages via the host-root base URL | Open guide |
| Terminal pair-programming | Aider with OPENAI_API_BASE + an openai/ model | Open guide |
| A single config-file IDE assistant | Continue via config.yaml roles | Open guide |
| A local multi-provider desktop chat | Cherry Studio as an OpenAI provider | Open guide |
| A self-hosted team chat UI | Open WebUI via an OpenAI connection | Open guide |
| The cheapest access to a given model | Start from the model-family guide | Open guide |
| To compare gateways honestly | DaoXE vs OpenRouter / LiteLLM / One-API | Open guide |
Verify us — don't trust us
Point the open benchmark at DaoXE and at the official API and compare at temperature 0. Then learn to detect model swapping so a cheaper endpoint can't quietly swap you to a smaller model.
Frequently asked questions#
What's the fastest way to start?
Find your tool in the editors list or the decision table, set the base URL to https://daoxe.com/v1 and paste your key. Most tools need nothing more.
Which tools support native Claude Code?
Claude Code uses native Anthropic Messages — see the Claude model guide and the Cursor & Claude Code guide. Point ANTHROPIC_BASE_URL at the host root.
How do I make sure I get the real model?
Benchmark the same prompt against the official API at temperature 0, and learn to detect model swapping. A cheaper endpoint is only a saving if it serves the real model.
Can one key drive several tools at once?
Yes — the same DaoXE key works across every integration here, on one balance. Current advantage: typically ~30–80% below list via low-price groups.
Try DaoXE — and benchmark it yourself
One key for GPT, Claude, Gemini, DeepSeek and more. Point the open benchmark at us and compare — don't take our word for it.