Use case · Multi-agent developers

One local gateway for every AI coding agent you use

Codex for one task, Claude Code for another, Gemini CLI or OpenCode on the side — and a different provider behind each. AutoJev gives them one endpoint, one set of providers and one log, while each agent keeps the models you pick for it.

One endpoint: http://127.0.0.1:9527/v1
One-click connect with config backup and restore
Different models or routes per agent
One log for every agent and provider

Download AutoJev Star on GitHub

Route every request to the right model.

Connect an agent

Codex

Claude Code

Gemini CLI

AutoJev

autojev/quality

openrouter/anthropic/claude-sonnet

deepseek/deepseek-chat

ollama/qwen3:14b

Providers

Models

Routes

Agents

Requests today

1,284

Tokens today

18.6M

Sound familiar?

Keys and URLs everywhere

The same provider key lives in ~/.codex/config.toml, ~/.claude/settings.json, an OpenCode JSON file and a Hermes YAML file. Rotating it is a scavenger hunt.

Every agent speaks a different API

Codex wants Responses, Claude Code wants Anthropic Messages, Gemini CLI wants the Gemini API. A provider that only serves Chat Completions is off limits to half of them.

Switching models means editing files

Trying a new model in Claude Code means rewriting ANTHROPIC_BASE_URL and ANTHROPIC_MODEL — and remembering to put them back.

No single view of usage

Each provider dashboard shows part of the picture. Nothing tells you which agent spent what on which model today.

How AutoJev fixes it

AutoJev is a free desktop app that runs one gateway on your machine. Providers and models are configured once, in one place. Each agent is connected to the gateway with one click: AutoJev backs up the agent’s config, writes the gateway URL and the models you selected, and restores the original when you disconnect.

The gateway accepts Chat Completions, Responses and Anthropic Messages, plus a Gemini bridge for Gemini CLI, and converts between them. So any agent can use any model you add — and every request ends up in the same local log with agent, model, provider, tokens, latency and estimated cost.

Agents → AutoJev → models

Agents talk to one local endpoint. AutoJev routes each request to a model and speaks the provider’s API.

Route every request to the right model.

Connect an agent

Codex

Claude Code

Gemini CLI

AutoJev

autojev/quality

openrouter/anthropic/claude-sonnet

deepseek/deepseek-chat

ollama/qwen3:14b

Providers

Models

Routes

Agents

Requests today

1,284

Tokens today

18.6M

Setup recipe: three agents, one gateway

About ten minutes from install to a shared log.

1
Install and add providers
Download AutoJev for macOS, Windows or Linux. Add providers on the Providers page — or import them from CC Switch or Termany, de-duplicated by endpoint. Test each provider.
2
Add models
Add the models you use with their upstream API type, prices, tier and capability flags. Each becomes available as `<provider>/<model_id>`.
3
Create a few routes
Group models into routes that match how you work, for example `autojev/fast` for quick edits, `autojev/quality` for reviews and `autojev/cheap` for bulk tasks. Pick load balancing or intelligent selection for each, or call a single model directly as <provider>/<model_id>.
text
autojev/fast → intelligent selection, Lower latency autojev/quality → intelligent selection, Higher quality autojev/cheap → load balancing across two inexpensive models
4
Connect each agent with its own models
On Agents, AutoJev lists the agents it detected. Select one or more models or routes per agent — Codex on `autojev/quality`, Hermes on `autojev/cheap`, Claude Code on both — and connect. Use Batch update to rewrite all connected agents after changes.
5
Watch it work
Open Logs. Each request shows the agent, route ID, the model that was chosen, tokens, first byte, total time, estimated cost and status.

Agents AutoJev connects today

Agent	Config AutoJev edits	Protocol
Codex	~/.codex/config.toml (+ model catalog)	Responses
Claude Code	~/.claude/settings.json	Anthropic Messages
Gemini CLI	~/.gemini/settings.json + .env	Gemini bridge
OpenCode	~/.config/opencode/opencode.json(c)	Chat or Messages
Hermes	~/.hermes/config.yaml	Chat
OpenClaw	~/.openclaw/openclaw.json	Chat or Messages
Kimi Code CLI	~/.kimi-code/config.toml	Chat or Messages
Grok Build	~/.grok/config.toml	Chat or Messages
oh-my-pi (OMP)	~/.omp/agent/models.yml, config.yml	Chat or Messages

Custom agents can be added with an executable and config path. Cursor CLI is detected but cannot be routed, because it does not expose a documented custom model endpoint setting.

Built for daily use

Original configs are backed up on connect and restored on disconnect, keeping later non-conflicting edits.
Closing the window keeps the gateway running in the tray; agents reconnect automatically on next launch.
Pause service rejects new requests while in-flight ones finish; Stop and restore puts every agent back.
Session affinity keeps a conversation on the same model while it stays eligible.
Failover retries another candidate on connection errors or 401/402/403/408/429/5xx before output starts.
Requests go straight from your machine to your providers. AutoJev is free and adds no markup.

Frequently asked questions

Can different agents use different models through AutoJev?

Yes. Model and route selection is per agent. Codex can be on one route and Claude Code on another, while both share the same providers.

What happens to my existing agent config?

AutoJev backs it up before writing anything and restores it when you disconnect, stop the gateway with restore, or quit from the tray. Non-conflicting edits you made later are kept.

Can Codex use a Claude model, or Claude Code use a GPT or DeepSeek model?

Yes. AutoJev converts between Responses, Chat Completions and Anthropic Messages in all directions, including streaming and function tools. Conversion cannot add capabilities a model lacks, and some features such as hosted tools are rejected.

Does AutoJev work with Cursor?

Cursor CLI is detected but not routed, because it has no documented setting for a custom model endpoint.

What if my agent is not on the list?

Add it as a custom agent, or use Connection parameters: base URL, placeholder key and model or route ID, with env, JSON and cURL snippets for OpenAI or Anthropic protocol.

Is AutoJev free?

Yes. AutoJev is free and open source under the MIT license. You pay your providers directly with your own keys.

Keep exploring

FeaturesOne-click agent setupConnect Codex, Claude Code, Gemini CLI, OpenCode and more. AutoJev edits each native config, backs it up and restores it.FeaturesProtocol conversionConvert between Chat Completions, Responses and Anthropic Messages in all six directions, with streaming and tool calls.Integrations

CodexRun Codex on DeepSeek, OpenRouter, Ollama or any model. AutoJev speaks Responses to Codex and converts for Chat-only providers.Integrations

Claude CodePoint Claude Code at AutoJev via ANTHROPIC_BASE_URL and use DeepSeek, GPT, Kimi or local models through the Messages API.FeaturesUsage & request logsTokens, latency and estimated cost for every request, broken down by agent, model and provider. Metadata only, stored locally.

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.