AutoJev
Download

Integration · Agent

Use DeepSeek, GPT or any model in Claude Code

Claude Code is Anthropic’s agentic coding tool and speaks the Anthropic Messages API. AutoJev connects it to a local gateway with one click, then routes every request to the provider and model you choose — converting Messages to Chat Completions or Responses when needed.

  • Sets ANTHROPIC_BASE_URL and model env for you
  • Your models replace the built-in picker options
  • Messages in, Chat or Responses out
  • Restores ~/.claude/settings.json on disconnect
Download AutoJevStar on GitHub

Agents

+ Agent

CodexResponses

~/.codex/config.toml

Connected

Claude CodeMessages

~/.claude/settings.json

Connected

Gemini CLIGemini

~/.gemini/settings.json

Connected

OpenCodeChat

~/.config/opencode/opencode.json

Connect

How the connection works

Claude Code reads its model endpoint from environment variables in ~/.claude/settings.json. When you connect it, AutoJev sets ANTHROPIC_BASE_URL to the local gateway, a placeholder ANTHROPIC_AUTH_TOKEN, and points ANTHROPIC_MODEL plus the default Haiku, Sonnet, Opus and Fable model variables at the model or route you chose as default.

Requests arrive at AutoJev as Anthropic Messages on POST /v1/messages. If the target model is Anthropic-compatible, they pass through. If it is a Chat Completions provider such as DeepSeek or an OpenAI-compatible host, or a Responses model, AutoJev converts the request and the streamed reply, including tool calls, images and usage.

Your real provider keys stay in AutoJev. Requests go straight from your machine to the provider — there is no AutoJev account or cloud relay in between.

What AutoJev configures

ItemValue
Config file~/.claude/settings.json
ProtocolAnthropic Messages (ANTHROPIC_BASE_URL = http://127.0.0.1:9527, no /v1)
AuthANTHROPIC_AUTH_TOKEN = autojev-local-claude (placeholder; labels requests as claude)
Default modelmodel + ANTHROPIC_MODEL and the DEFAULT_*_MODEL variables
Model pickermodelPicker.options lists every model and route you selected

Other keys in settings.json, including your own env entries, are left untouched.

Set up Claude Code with AutoJev

  1. 1

    Install AutoJev

    Download the desktop app and open it. The gateway listens on http://127.0.0.1:9527 on localhost only.

  2. 2

    Add a provider and models

    Use a preset (DeepSeek, OpenRouter, OpenAI, Ollama) or any OpenAI-compatible or Anthropic-compatible endpoint. Add your key, test the provider, and add the models you want with their upstream API type.

  3. 3

    Optional: create a route

    A route such as autojev/fast or autojev/cheap lets AutoJev pick the model per request — load balancing, or intelligent selection by cost, quality or speed. To pin one model, call <provider>/<model_id> directly.

  4. 4

    Connect Claude Code

    Open Agents, select one or more models or routes for Claude Code, choose the default, and click Connect.

  5. 5

    Start Claude Code

    Open a new session. Use /model to switch between the options AutoJev added.

    bash
    claude

The settings AutoJev writes

With autojev/fast as the default, AutoJev merges these keys into settings.json. One modelPicker option is added per selected model or route.

~/.claude/settings.json
{ "model": "autojev/fast", "modelPicker": { "replaceBuiltInOptions": true, "options": [ { "model": "autojev/fast", "label": "Fast", "description": "AutoJev · autojev/fast", "behavesAs": "claude-sonnet-4-6" } ] }, "env": { "ANTHROPIC_BASE_URL": "http://127.0.0.1:9527", "ANTHROPIC_AUTH_TOKEN": "autojev-local-claude", "ANTHROPIC_MODEL": "autojev/fast", "ANTHROPIC_DEFAULT_HAIKU_MODEL": "autojev/fast", "ANTHROPIC_DEFAULT_SONNET_MODEL": "autojev/fast", "ANTHROPIC_DEFAULT_OPUS_MODEL": "autojev/fast", "ANTHROPIC_DEFAULT_FABLE_MODEL": "autojev/fast" } }

Switching models and routes

Every model and route you selected appears in Claude Code’s model picker, and AutoJev replaces the built-in options so the list only shows what the gateway can serve. The behavesAs field is a client-side compatibility hint only; it never changes which upstream model AutoJev calls.

To change the default or the list, edit the selection on the Agents page and reconnect, or use “Batch update configurations” for all connected agents. If you connected a route, editing the route in AutoJev is enough — Claude Code keeps sending autojev/<route-id>.

Restoring settings.json

Disconnect, “Stop and restore configurations”, or “Quit AutoJev (restore agents)” revert the keys AutoJev set. A field you changed yourself after connecting is kept, and unrelated env entries are never touched. AutoJev reconnects on the next launch unless you disconnected manually.

Limits to know

  • Thinking blocks are dropped when a request is converted to another protocol.
  • Conversion cannot give a model tools, vision or reasoning it does not have — choose tool-capable models for coding.
  • Hosted tools and audio or file content are rejected when converting.
  • Claude Code only works through AutoJev while the gateway is running.

Frequently asked questions

How do I use DeepSeek in Claude Code?

Add the DeepSeek preset in AutoJev, add a DeepSeek model, then connect Claude Code on the Agents page. AutoJev converts Claude Code’s Messages requests to the model’s API type and back.

Do I still need an Anthropic API key?

Not for models served through AutoJev. Claude Code sends the placeholder token autojev-local-claude to the local gateway, and AutoJev uses the provider key you added for the chosen model.

Why does ANTHROPIC_BASE_URL have no /v1?

Anthropic clients append /v1/messages themselves, so the base URL is http://127.0.0.1:9527.

Can I switch models inside Claude Code?

Yes. AutoJev lists every model and route you selected in the model picker. Restart Claude Code after changing the selection so it reloads the list.

Is this the same as editing settings.json by hand?

The result is similar, but AutoJev backs up the original, tracks which keys it owns, and restores them on disconnect or quit while keeping your later edits.

Can Claude Code use a local model?

Yes. Add the Ollama preset or any local OpenAI-compatible server, add a tool-capable model, and connect. Routes can also prefer local models.

Keep exploring

FeaturesProtocol conversionConvert between Chat Completions, Responses and Anthropic Messages in all six directions, with streaming and tool calls.IntegrationsDeepSeekRun DeepSeek in Codex and Claude Code. AutoJev converts Responses and Messages to Chat Completions on localhost.CompareClaude Code Router alternativeConnect Claude Code to DeepSeek, OpenAI, OpenRouter or local models in one click, with load balancing and failover.FeaturesOne-click agent setupConnect Codex, Claude Code, Gemini CLI, OpenCode and more. AutoJev edits each native config, backs it up and restores it.Use casesFor multi-agent developersCodex, Claude Code, Gemini CLI, OpenCode and more on one local gateway, with per-agent models and one request log.

Updated 2026-10-14

Give every agent the right model.

Install AutoJev, add the providers you already pay for, and connect your agents in one click. Everything stays on your machine.

Download AutoJevStar on GitHub

Free & open source · macOS · Windows · Linux