Provider integration

Use any Anthropic-compatible API with every coding agent

Many providers now offer an Anthropic Messages endpoint so they can plug into Claude Code. AutoJev takes that one endpoint and makes it available to Codex, Gemini CLI, OpenCode, Hermes and the rest too.

Anthropic Compatible preset with API type Anthropic Messages
Native passthrough for Claude Code
Messages → Responses / Chat conversion for other agents
Switch Claude Code between providers without editing settings.json

Download AutoJev Star on GitHub

Providers

+ Providers

OpenRouter

https://openrouter.ai/api

DeepSeek

https://api.deepseek.com

OpenAI

https://api.openai.com/v1

Ollama

http://127.0.0.1:11434

Stop rewriting ANTHROPIC_BASE_URL by hand

The usual way to run Claude Code on another provider is to edit ~/.claude/settings.json: change ANTHROPIC_BASE_URL, swap the token, set ANTHROPIC_MODEL, and remember to change it all back. Doing that for several providers — and for every other agent you use — does not scale.

With AutoJev, Claude Code is connected once to the local gateway. Anthropic-compatible providers are added once in AutoJev with the Anthropic Messages API type. Switching providers becomes a model or route choice, and the same providers are available to every other connected agent through protocol conversion.

Add an Anthropic-compatible provider

1
Choose Anthropic Compatible
Providers → Add provider → Anthropic Compatible (under Custom). The API type is set to Anthropic Messages. Enter the provider’s base URL — the part before /v1/messages — and your key.
text
Provider type: Anthropic Compatible Provider name: deepseek-anthropic Base URL: https://api.deepseek.com/anthropic API type: Anthropic Messages API key: sk-…
2
Test the endpoint
Enter a model ID and click Test. AutoJev calls <base URL>/v1/messages with the x-api-key header and anthropic-version 2023-06-01, and shows the exact endpoint it used.
3
Add models
Add each model with API type Anthropic Messages, its model ID, prices per 1M tokens, tier, capability flags and context window.
4
Connect Claude Code — and everything else
Select the model or a route on the Agents page and connect. Claude Code reaches it natively; Codex, OpenCode, Hermes and the other agents reach it through conversion.

Anthropic-compatible endpoints you can add

Any endpoint that implements Anthropic Messages works, such as these. URLs are from each provider’s documentation and may differ by region.

Provider	Base URL
Anthropic	https://api.anthropic.com
DeepSeek	https://api.deepseek.com/anthropic
GLM / Zhipu (China)	https://open.bigmodel.cn/api/anthropic
GLM / Z.ai (international)	https://api.z.ai/api/anthropic
MiniMax (international / China)	https://api.minimax.io/anthropic · https://api.minimaxi.com/anthropic

AutoJev sends the key as the x-api-key header on Messages requests. If a provider only accepts a Bearer token, the provider test will show the upstream error.

What Claude Code is configured with

When you connect Claude Code, AutoJev backs up ~/.claude/settings.json and sets these env entries. The token is a local placeholder; your provider key stays in AutoJev.

~/.claude/settings.json (excerpt)

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://127.0.0.1:9527",
    "ANTHROPIC_AUTH_TOKEN": "autojev-local-claude",
    "ANTHROPIC_MODEL": "autojev/quality",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "autojev/quality",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "autojev/quality",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "autojev/quality"
  }
}

How each agent reaches a Messages provider

Agent	Agent speaks	Path
Claude Code	Anthropic Messages	Native passthrough
Codex	Responses	Responses → Messages, converted back
Hermes	Chat Completions	Chat → Messages, converted back
OpenCode, Kimi, Grok Build, OpenClaw, OMP	Chat or Messages	Native when the agent uses Messages
Gemini CLI	Gemini bridge	Bridge → Chat → Messages

Anthropic-compatible or OpenAI-compatible?

Some providers offer both. You can add the same provider twice under different provider names and compare them.

Choose Anthropic Messages when…

Claude Code is your main agent and you want native passthrough, including anthropic-beta headers, or the provider’s OpenAI-style path does not fit AutoJev’s /v1 URL scheme (as with GLM).

Choose Chat Completions when…

Most of your traffic comes from Chat-speaking agents such as Hermes or OpenCode, or you rely on structured output, which converts between Chat Completions and Responses only.

Thinking blocks do not cross protocols

On native Messages requests from Claude Code, provider-specific fields such as anthropic-beta headers are forwarded. When a request is converted for another agent, thinking blocks and signatures are dropped and reasoning controls are not translated.

Frequently asked questions

What base URL do I enter for an Anthropic-compatible provider?

The URL that Claude Code would use as ANTHROPIC_BASE_URL — everything before /v1/messages. AutoJev appends /v1/messages itself.

Can Codex use an Anthropic-only provider?

Yes. Codex sends Responses to AutoJev, which converts the request, tool history and stream to and from Anthropic Messages.

Can I switch Claude Code between providers without editing settings.json?

Yes. Claude Code stays pointed at the gateway. Change the models in its route, or reconnect it with another model or route, and AutoJev updates the config for you.

Is an Anthropic preset included?

There is no dedicated Anthropic preset. Use Anthropic Compatible with https://api.anthropic.com as the base URL.

Are beta headers forwarded?

Only on unchanged protocols: when Claude Code calls a Messages provider natively, anthropic-version and anthropic-beta headers are passed through.