Kernl

Connect a model

Until you connect an AI, Kernl's chat and agents stay off. Everything else works. You connect a model in the dashboard.

Open the AI settings

In the sidebar choose Settings, then AI. You can also go straight to http://localhost:3086/settings?section=ai.

The page is called Connect your AI. Press Connect provider to pick one. Providers are grouped as Free, On your computer, Pay as you go and With your subscription.

Use a cloud API key

  1. Pick a provider card, for example NVIDIA (marked Recommended and free), Google Gemini, Groq, Anthropic or OpenAI.
  2. Click the link that opens the provider's site and create an API key there.
  3. Paste it into the API key field.
  4. Click Connect.

When it works you see "Connected" with a response time, and a Model list to choose from. If the key looks wrong, Kernl warns you but lets you try it anyway.

Use a local model

Ollama and LM Studio run on your own machine and need no key.

For Ollama: install it from ollama.com, download a model with the command below, and leave Ollama running.

ollama pull qwen3

For LM Studio: install it from lmstudio.ai, download and load a model, then in Developer turn on Start server.

Back in Kernl, pick the provider and click Detect. The Server address is prefilled with the usual one. When Kernl finds your model it shows "Found at ..." and you click Connect.

If you run Kernl in Docker, localhost inside the container is not your computer. The stack's compose file maps host.docker.internal to your host, so use it in the server address, for example http://host.docker.internal:11434/v1 for Ollama. We read this from the compose file and have not tried it (not run by us).

Use your Claude Code subscription

You can let Kernl use your Claude subscription as its AI. This is the opposite of connecting Claude Code to Kernl, where Claude Code uses Kernl's tools. The two are unrelated and you can do both.

  1. Under With your subscription, pick Claude Code.
  2. Click Connect with my subscription. A Claude approval page opens in your browser.
  3. Approve it. You are done.

In Docker the browser cannot hand the approval back by itself. Click Open the approval page, approve, and your browser lands on a page that will not load. Its address starts with http://localhost. Copy that whole address, paste it into the field Address of the page that didn't load, and click Connect.

If the one-button flow fails, use Or sign in from a terminal and run the command it shows.

Next step

To use Kernl's tools from another app, read MCP basics.