Provider integration

Run Ollama local models in any coding agent

Point AutoJev at your Ollama server once. Every connected agent can then use your local models through the same gateway it uses for cloud models — and routes can prefer local whenever a local model is good enough.

Built-in preset: http://127.0.0.1:11434, no API key required
"Prefer local models" routing option
Mix local and cloud models with failover
Gateway and logs stay on your machine

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

Local models, without per-agent wiring

Ollama makes running models locally easy. Getting every coding agent to use them is the annoying part: Codex expects a Responses endpoint, Claude Code expects Anthropic Messages, and each agent has its own config format. AutoJev handles both problems. It connects agents to one localhost gateway and converts their requests to the Chat Completions API that Ollama serves.

Because the AutoJev gateway listens on localhost only and request logs keep metadata only, a request from your agent to a local Ollama model never leaves your machine.

Add Ollama

1
Make sure Ollama is running
Start Ollama and pull the models you want to use. Note the exact model names that `ollama list` prints — you will enter them as model IDs.
bash
ollama pull qwen3-coder:30b ollama list
2
Create the provider
Providers → Add provider → Ollama. The base URL is preset to http://127.0.0.1:11434 and no API key is needed. Change the URL if Ollama runs on another port or machine.
text
Provider type: Ollama Provider name: ollama Base URL: http://127.0.0.1:11434 API type: OpenAI Chat Completions API key: (not required)
3
Test and add models
Test with one of your model names; the request goes to http://127.0.0.1:11434/v1/chat/completions. Then add each model with its tier, capability flags and context window. Leave prices at zero if you only want to track tokens. Agents see `ollama/qwen3-coder:30b`.
4
Run a speed test
On Models, run a speed test. AutoJev sends three short streaming requests per model and records first-content latency and output tokens/s, so you know how your hardware compares with cloud models.

Routing options that involve local models

Setting	What it does
Single model	Call one Ollama model directly as <provider>/<model_id>, or give a load-balancing route a single Ollama candidate.
Load balancing	Spread traffic across local and cloud models by weight, with priority groups (higher numbers first).
Intelligent selection + Prefer local models	Choose Ollama when capability and quality tiers are equivalent to the cloud candidates.
Intelligent selection + Lower latency	Pick by measured speed. The local preference is hidden here because measurements decide.
Failover	If a candidate fails with a connection error or 401/402/403/408/429/5xx before output starts, try the next one.

A local-first route in practice

Create a route called `local` with an Ollama model at a higher priority and a cloud model as backup, then call it like any model.

Calling a local-first route

curl http://127.0.0.1:9527/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer autojev-local-terminal' \
  -d '{"model": "autojev/local", "messages": [{"role": "user", "content": "Write a unit test for parseDate"}]}'

What to expect

Every agent, same models

Codex, Claude Code, Gemini CLI, OpenCode, Hermes and the rest all see your Ollama models after one connection each.

Honest capability flags

Set tools, vision and reasoning flags to match the model. Conversion cannot add capabilities a local model does not have.

Usage stays visible

Logs and Usage show local requests next to cloud ones, broken down by agent, model and provider.

Local hardware has limits

Coding agents send long contexts and many tool calls. Set the context window you actually run Ollama with, and keep a cloud model in the route as a fallback for requests a local model cannot handle.

Frequently asked questions

Does Ollama need an API key in AutoJev?

No. The Ollama provider type requires no API key; AutoJev calls your Ollama server directly.

Can Codex or Claude Code use Ollama models through AutoJev?

Yes. Codex sends Responses and Claude Code sends Anthropic Messages to the gateway; AutoJev converts both to Chat Completions for Ollama and converts the stream back.

What does "Prefer local models" do?

In intelligent selection routes, it makes AutoJev choose an Ollama model when its capability and quality tier are equivalent to the other candidates. It is hidden for the Lower latency preference, which ranks purely by measured speed.

Do my prompts leave my machine when I use Ollama?

Not through AutoJev: the gateway listens on localhost, the request goes to your Ollama server, and logs store metadata only. If you use Jev for a Balanced or Higher quality route, routing metadata — not conversation content — is sent to OpenRouter.

Ollama runs on another machine. Does that work?

Yes. Change the base URL to that machine’s address and port. Speed tests will then include the network latency.

Keep exploring

Use casesFor local-first developersPrefer Ollama models, fail over to the cloud when needed. Localhost gateway, metadata-only logs, no account.FeaturesSmart model routingExpose routes like autojev/fast to every agent. Load-balance models or let Jev pick by cost, quality, speed or balance.FeaturesModel speed testsMeasure first-content latency and output tokens per second on your own network, on a schedule, and route by the results.FeaturesFailover & load balancingWeighted pools, priority groups and sticky sessions, with automatic failover on 429s and 5xx errors and Retry-After cooldowns.Integrations

OpenAI-compatible APIsKimi, Qwen, SiliconFlow, MiniMax, vLLM, LM Studio — any OpenAI-compatible endpoint, available to all your agents.

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.