Use Kimi in Claude Code

Claude Code is a programming agent product from Anthropic. Its interface, configuration options, and supported capabilities may change across versions. This guide describes a general integration approach: forwarding Claude Code model requests to the Kimi API through environment variables.

Install Claude Code

Skip this step if Claude Code is already installed. Run:

npm install -g @anthropic-ai/claude-code

No Node.js yet? Expand for installation and initialization

macOS and Linux:

# Install Node.js
curl -fsSL https://fnm.vercel.app/install | bash

# Open a new terminal so fnm takes effect
fnm install 24.3.0
fnm default 24.3.0
fnm use 24.3.0

Windows (PowerShell):

# Right-click the Windows button, click "Terminal", then run:
winget install OpenJS.NodeJS
Set-ExecutionPolicy -Scope CurrentUser RemoteSigned

# Close the terminal window and open a new one

After installing Node.js, run the initialization once:

node --eval "
    const fs = require('fs');
    const path = require('path');
    const os = require('os');
    const homeDir = os.homedir();
    const filePath = path.join(homeDir, '.claude.json');
    if (fs.existsSync(filePath)) {
        const content = JSON.parse(fs.readFileSync(filePath, 'utf-8'));
        fs.writeFileSync(filePath, JSON.stringify({ ...content, hasCompletedOnboarding: true }, null, 2), 'utf-8');
    } else {
        fs.writeFileSync(filePath, JSON.stringify({ hasCompletedOnboarding: true }, null, 2), 'utf-8');
    }"

Not a fresh install? Clean up legacy configuration and environment variables

If you previously modified ~/.claude/settings.json through third-party tools or by hand, stale entries in its env field override environment variables exported in your terminal, which can silently prevent the new configuration from taking effect or rewrite model requests. Run the following script to clean them up first:

node --eval "
    const fs = require('fs');
    const path = require('path');
    const os = require('os');
    const settingsPath = path.join(os.homedir(), '.claude', 'settings.json');
    if (fs.existsSync(settingsPath)) {
        const content = JSON.parse(fs.readFileSync(settingsPath, 'utf-8'));
        if (content && typeof content === 'object' && content.env && typeof content.env === 'object') {
            for (const key of [
                'ANTHROPIC_BASE_URL',
                'ANTHROPIC_API_KEY',
                'ANTHROPIC_AUTH_TOKEN',
                'ANTHROPIC_MODEL',
                'ANTHROPIC_SMALL_FAST_MODEL',
                'CLAUDE_CODE_SUBAGENT_MODEL',
                'ANTHROPIC_DEFAULT_OPUS_MODEL',
                'ANTHROPIC_DEFAULT_OPUS_MODEL_NAME',
                'ANTHROPIC_DEFAULT_SONNET_MODEL',
                'ANTHROPIC_DEFAULT_SONNET_MODEL_NAME',
                'ANTHROPIC_DEFAULT_HAIKU_MODEL',
                'ANTHROPIC_DEFAULT_HAIKU_MODEL_NAME',
                'ANTHROPIC_DEFAULT_FABLE_MODEL',
                'ANTHROPIC_DEFAULT_FABLE_MODEL_NAME',
                'ENABLE_TOOL_SEARCH',
                'CLAUDE_CODE_AUTO_COMPACT_WINDOW',
                'CLAUDE_CODE_EFFORT_LEVEL',
            ]) {
                delete content.env[key];
            }
            fs.writeFileSync(settingsPath, JSON.stringify(content, null, 2), 'utf-8');
        }
    }"

This script only removes endpoint, credential, and model-related variables from env; other settings (such as permissions and theme) are left untouched.Also check shell configuration files such as ~/.zshrc and ~/.bashrc for stale ANTHROPIC_* exports (on Windows, check your user environment variables) and remove any you find — they interfere with the new configuration just as much.

Get a Kimi API Key

Visit Kimi Open Platform to create an API key (choose the default project), and use it in place of YOUR_MOONSHOT_API_KEY below.

Configure Environment Variables

Choose one of the two methods below and do not mix them: Method 1 takes effect immediately but only lasts for the current terminal session; Method 2 writes to a configuration file and persists.

Method 1: Terminal Environment Variables (Current Session Only)

macOS and Linux:

export ANTHROPIC_BASE_URL="https://api.moonshot.ai/anthropic"
export ANTHROPIC_AUTH_TOKEN="${YOUR_MOONSHOT_API_KEY}"
export ANTHROPIC_MODEL="kimi-k3[1m]"
export ANTHROPIC_DEFAULT_OPUS_MODEL="kimi-k3[1m]"
export ANTHROPIC_DEFAULT_SONNET_MODEL="kimi-k3[1m]"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="kimi-k3[1m]"
export ANTHROPIC_DEFAULT_FABLE_MODEL="kimi-k3[1m]"
export CLAUDE_CODE_SUBAGENT_MODEL="kimi-k3[1m]"
export CLAUDE_CODE_AUTO_COMPACT_WINDOW="1048576"
export CLAUDE_CODE_EFFORT_LEVEL="max"
claude

Windows (PowerShell):

$env:ANTHROPIC_BASE_URL="https://api.moonshot.ai/anthropic";
$env:ANTHROPIC_AUTH_TOKEN="YOUR_MOONSHOT_API_KEY"
$env:ANTHROPIC_MODEL="kimi-k3[1m]"
$env:ANTHROPIC_DEFAULT_OPUS_MODEL="kimi-k3[1m]"
$env:ANTHROPIC_DEFAULT_SONNET_MODEL="kimi-k3[1m]"
$env:ANTHROPIC_DEFAULT_HAIKU_MODEL="kimi-k3[1m]"
$env:ANTHROPIC_DEFAULT_FABLE_MODEL="kimi-k3[1m]"
$env:CLAUDE_CODE_SUBAGENT_MODEL="kimi-k3[1m]"
$env:CLAUDE_CODE_AUTO_COMPACT_WINDOW="1048576"
$env:CLAUDE_CODE_EFFORT_LEVEL="max"
claude

Method 2: Write To settings.json (Persistent)

Write the same variables into the env field of ~/.claude/settings.json:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.moonshot.ai/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_MOONSHOT_API_KEY",
    "ANTHROPIC_MODEL": "kimi-k3[1m]",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "kimi-k3[1m]",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "kimi-k3[1m]",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "kimi-k3[1m]",
    "ANTHROPIC_DEFAULT_FABLE_MODEL": "kimi-k3[1m]",
    "CLAUDE_CODE_SUBAGENT_MODEL": "kimi-k3[1m]",
    "CLAUDE_CODE_AUTO_COMPACT_WINDOW": "1048576",
    "CLAUDE_CODE_EFFORT_LEVEL": "max"
  }
}

Note: env values in settings.json override variables exported in your terminal; this file contains your API key in plaintext — do not commit it to a git repository; restart Claude Code after saving.

Configuration Reference

Claude Code uses different model tiers for different scenarios (main conversation, background summarization, sub-agents, and so on). Configuring only some of the variables makes the corresponding scenarios fail silently:

Variable	Purpose	Impact if missing or misconfigured
`ANTHROPIC_BASE_URL`	Forwards model requests to Kimi’s Anthropic-compatible endpoint	Requests go to Anthropic’s official endpoint and fail authentication
`ANTHROPIC_AUTH_TOKEN`	Authenticates with your Kimi API key	Returns 401 authentication errors
`ANTHROPIC_MODEL`	Model used for the main conversation	Falls back to a default Claude model name that the Kimi endpoint cannot recognize, resulting in model-not-found errors
`ANTHROPIC_DEFAULT_OPUS_MODEL` / `ANTHROPIC_DEFAULT_SONNET_MODEL` / `ANTHROPIC_DEFAULT_HAIKU_MODEL` / `ANTHROPIC_DEFAULT_FABLE_MODEL`	Model names Claude Code uses when selecting models by task tier	Tasks on the corresponding tier (e.g. background title generation and summarization on the haiku tier) fail
`CLAUDE_CODE_SUBAGENT_MODEL`	Model used by sub-agents	Sub-agent tasks fail or degrade noticeably
`CLAUDE_CODE_AUTO_COMPACT_WINDOW`	Context window size that triggers automatic compaction	Must match the model’s context: `1048576` for `kimi-k3` (1M), `262144` for `kimi-k2.7-code` (256K). Too small compacts prematurely and loses context; too large causes context-length errors
`CLAUDE_CODE_EFFORT_LEVEL`	Controls Claude Code’s reasoning effort	Set to `max` to enable the most thorough reasoning; lower values may reduce quality for complex tasks

Models And Thinking Behavior

How the three models actually behave in Claude Code (verified against the Anthropic-compatible endpoint):

Model	Thinking	Notes
`kimi-k3` (default on this page)	On by default	Works out of the box, no extra configuration needed
`kimi-k2.7-code`	Always on	Requests must explicitly enable thinking — keep Thinking on (press `Tab`) in Claude Code. With thinking off, requests are rejected (`invalid thinking: only type=enabled is allowed for this model`)
`kimi-k2.6`	Optional	Can run with thinking off; good for latency-sensitive simple tasks

When switching models, replace the value of every model variable in the configuration with the new model name.

Confirm That The Configuration Took Effect

In Claude Code, enter /status to confirm the configuration:

Base URL should show https://api.moonshot.ai/anthropic
Model should show kimi-k3[1m]

Claude Code’s /model menu is a built-in list of fixed aliases — it does not show Kimi models, and you do not need to switch anything there. Whether the configuration took effect is determined by what /status shows.

Finally, send any message (for example hi). Receiving a normal reply confirms the end-to-end setup works.

Turn On Thinking

kimi-k3 thinks by default and works out of the box. If you switch to kimi-k2.7-code, it requires requests to explicitly enable thinking: press Tab in Claude Code to turn Thinking on and start working once you see the “Thinking on” indicator — otherwise the model rejects requests (400 invalid thinking), and features such as WebSearch do not work.

You can now use Claude Code for development normally.

Switch To The High-Speed Model

Kimi K2.7 Code offers a high-speed variant, kimi-k2.7-code-highspeed, with an output speed about 5-6x that of the regular version. If you prioritize output speed, change the value of every model variable in the configuration to kimi-k2.7-code-highspeed (note that it requires explicitly enabled thinking — see Models And Thinking Behavior). See K2.7 Code pricing for details.

Third-Party Tool: cc-switch

Community tools such as cc-switch can switch between multiple provider configurations. These tools are not maintained by Kimi, and their presets may differ from the values recommended on this page. After using them, verify each variable against the Configuration Reference above, and use /status to confirm the Base URL and model actually in effect.

FAQ

WebSearch fails with 400 invalid thinking / WebFetch unavailable

WebSearch fails with 400 invalid thinking: only type=enabled is allowed for this model: kimi-k2.7-code requires thinking, and WebSearch requests without it are rejected by the platform. Press Tab to turn Thinking on first, then retry; if it still fails, switch to kimi-k2.6 (thinking optional, not subject to this restriction). kimi-k3 is not affected. This is unrelated to your local configuration or cc-switch.
WebFetch reports temporarily unavailable or returns no fetched content: the endpoint does not support WebFetch for now — unrelated to your configuration; it will work once the platform adds support. As a workaround, paste the web page content into the chat, or use an MCP scraping tool instead.

401 authentication errors

Check that ANTHROPIC_AUTH_TOKEN is a valid Kimi API key. If you previously configured ANTHROPIC_API_KEY, remove it to avoid conflicts with ANTHROPIC_AUTH_TOKEN when both are present.

Model not found

Check the spelling of every model variable (kimi-k3[1m]), and make sure there are no extra spaces or quotes.

Background tasks or sub-agents fail

This usually means ANTHROPIC_DEFAULT_HAIKU_MODEL, ANTHROPIC_DEFAULT_FABLE_MODEL, or CLAUDE_CODE_SUBAGENT_MODEL is not set, so those scenarios requested a model name the Kimi endpoint cannot recognize. Fill in the missing variables per the Configuration Reference above.

Configuration changes do not take effect

Check for stale entries in the env field of ~/.claude/settings.json (they override terminal environment variables); you can run the cleanup script in the collapsed section above;
Variables exported in a terminal only apply to the current session and must be set again in a new window. If you appended them to ~/.zshrc or used the settings.json method, make sure you restarted Claude Code after the change.

API key and endpoint mismatch

Make sure ANTHROPIC_BASE_URL matches the platform where you created the API key — create the key on the platform linked in “Get a Kimi API Key” above and use the endpoint shown on this page.

Get Started

Model Capabilities

Tooling workflows

Core workflows

Integrations

Debugging and operations

Install Claude Code

Get a Kimi API Key

Configure Environment Variables

Method 1: Terminal Environment Variables (Current Session Only)

Method 2: Write To settings.json (Persistent)

Configuration Reference

Models And Thinking Behavior

Confirm That The Configuration Took Effect

Turn On Thinking

Switch To The High-Speed Model

Third-Party Tool: cc-switch

FAQ

​Install Claude Code

​Get a Kimi API Key

​Configure Environment Variables

​Method 1: Terminal Environment Variables (Current Session Only)

​Method 2: Write To settings.json (Persistent)

​Configuration Reference

​Models And Thinking Behavior

​Confirm That The Configuration Took Effect

​Turn On Thinking

​Switch To The High-Speed Model

​Third-Party Tool: cc-switch

​FAQ

Install Claude Code

Get a Kimi API Key

Configure Environment Variables

Method 1: Terminal Environment Variables (Current Session Only)

Method 2: Write To settings.json (Persistent)

Configuration Reference

Models And Thinking Behavior

Confirm That The Configuration Took Effect

Turn On Thinking

Switch To The High-Speed Model

Third-Party Tool: cc-switch

FAQ