Use Kimchi Router in Claude Code | Kimchi Inference

This guide shows you how to run Claude Code, Anthropic's agentic coding tool, on the Kimchi Model Router. Tested with Claude Code 2.1.277.

Once connected, Claude Code sends its requests to the auto model, and the Model Router picks the model that fits each task.

Prerequisites

The Kimchi CLI installed and set up. See Kimchi CLI.
Claude Code 2.1.242 or later.

Connect Claude Code to Kimchi

Run kimchi setup-tools

Run the setup wizard and select Claude Code:

kimchi setup-tools

The wizard points Claude Code at the Kimchi gateway and attaches your credentials. For a manual setup with a settings.json file, see the Claude Code recipe instead.

Start a session on auto

Start Claude Code with the auto model:

claude --model auto

Verify the connection

Ask a simple question, then open the Kimchi console. Go to Analytics and set the time range to 24h. Your requests appear in the totals within seconds.

To see which models the Router picked, see Router analytics.

Context window

Claude Code plans a 200K-token window for model IDs it does not recognize, and auto is one of them.

To plan for the full 1M window, start with:

claude --model 'auto[1m]'

The [1m] suffix is a local hint. Claude Code strips it before the request is sent. The Model Router may still pick a model with a smaller window, such as Kimi K3 at 262K. When that happens, requests can fail once the conversation passes the real limit of the picked model.

Switching models

Run /model auto in a session to switch to the Router. Claude Code saves the choice as the default for future sessions.

The /model dropdown does not list auto by default. To pin it into the picker with a label, add a modelPicker entry to ~/.claude/settings.json. The Claude Code recipe shows the format.

Supported features

Image support

Limitations and known issues

Claude desktop app

The Claude desktop app, including its Code tab, cannot use Kimchi models. Its model list accepts Anthropic model names only. Run Claude Code in the terminal or through the VS Code or JetBrains extensions instead.

Next steps

Kimchi Model Router

How routing works, pricing, and Router analytics.

Claude Code recipe

The full Claude Code gateway setup, including manual configuration.