How to Set Up Claude Code with the MiniMax API in 2026

Claude Code is Anthropic’s official command-line coding agent. It’s powerful, well-designed, and — until recently — locked to Anthropic’s own API. Which meant you had to pay Anthropic pricing, route your code through their servers, and trust the defaults.

You can now point Claude Code at any OpenAI-compatible API endpoint. That includes OpenAI, DeepSeek, MiniMax, OpenRouter, or any other provider that exposes the chat-completions interface. Run the same Claude Code interface, but with the model that makes sense for you.

This guide covers the full setup — from a clean machine to a working Claude Code that talks to the MiniMax API. Works on macOS, Linux, and Windows.

Why MiniMax Over the Default Anthropic API?

Reason	Details
:—	:—
Cost	MiniMax’s promo pricing is a fraction of Anthropic’s API rates
Performance	M3 scores competitively with Claude on coding benchmarks
Data sovereignty	Your code goes to a Chinese lab instead of US-based Anthropic
Reasoning capability	M3 is strong on multi-step agent tasks — exactly what Claude Code does

For solo developers and small teams, this is the cheapest way to use Claude Code.

What You’ll Need

– A machine with Node.js 18+ (any OS)

– A MiniMax API key (and optionally a Group ID for TTS)

– 10 minutes

Step 1: Install Claude Code

The official way:

“`bash

macOS / Linux / WSL

curl -fsSL https://claude.ai/install.sh | sh

Or via npm (works everywhere)

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

Windows (PowerShell)

irm https://claude.ai/install.ps1 | iex

“`

Verify:

“`bash

claude –version

Should print something like 2.1.x

“`

Step 2: Get Your API Credentials

You need a MiniMax API key. The endpoint format is `https://api.minimaxi.com/anthropic` (note the path — it’s the Anthropic-compatible endpoint, not the standard OpenAI-compatible one).

Get your key from the MiniMax platform dashboard. You’ll have:

– `ANTHROPIC_API_KEY` — your MiniMax API key

– `ANTHROPIC_BASE_URL` — set to `https://api.minimaxi.com/anthropic`

– `ANTHROPIC_MODEL` — the model name (e.g., `MiniMax-M3`, `MiniMax-M2.7`, `MiniMax-M2.7-highspeed`)

Step 3: Configure Claude Code

Claude Code reads environment variables. Set them in your shell profile so they persist:

On macOS / Linux / WSL — add to `~/.bashrc`, `~/.zshrc`, or `~/.profile`:

“`bash

export ANTHROPIC_BASE_URL=”https://api.minimaxi.com/anthropic”

export ANTHROPIC_API_KEY=”your-minimax-api-key-here”

export ANTHROPIC_MODEL=”MiniMax-M3″

export ANTHROPIC_DEFAULT_SONNET_MODEL=”MiniMax-M3″

export ANTHROPIC_DEFAULT_OPUS_MODEL=”MiniMax-M3″

export ANTHROPIC_SMALL_FAST_MODEL=”MiniMax-M3″

export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1

“`

On Windows (PowerShell) — run in PowerShell as admin:

“`powershell

[System.Environment]::SetEnvironmentVariable(“ANTHROPIC_BASE_URL”, “https://api.minimaxi.com/anthropic”, “User”)

[System.Environment]::SetEnvironmentVariable(“ANTHROPIC_API_KEY”, “your-minimax-api-key-here”, “User”)

[System.Environment]::SetEnvironmentVariable(“ANTHROPIC_MODEL”, “MiniMax-M3”, “User”)

[System.Environment]::SetEnvironmentVariable(“ANTHROPIC_DEFAULT_SONNET_MODEL”, “MiniMax-M3”, “User”)

[System.Environment]::SetEnvironmentVariable(“ANTHROPIC_DEFAULT_OPUS_MODEL”, “MiniMax-M3”, “User”)

[System.Environment]::SetEnvironmentVariable(“ANTHROPIC_SMALL_FAST_MODEL”, “MiniMax-M3”, “User”)

[System.Environment]::SetEnvironmentVariable(“CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC”, “1”, “User”)

“`

Reload the shell:

“`bash

source ~/.bashrc # or: exec zsh # or: open a new PowerShell window

“`

Step 4: Skip the Onboarding (Optional)

Claude Code’s first run normally asks you to log in to an Anthropic account. Skip this by pre-creating the config file:

On macOS / Linux:

“`bash

mkdir -p ~/.claude

cat > ~/.claude.json << 'EOF'

{

“hasCompletedOnboarding”: true

}

EOF

“`

On Windows (PowerShell):

“`powershell

New-Item -ItemType Directory -Path “$env:USERPROFILE\.claude” -Force

@’

{

“hasCompletedOnboarding”: true

}

‘@ | Set-Content “$env:USERPROFILE\.claude.json”

“`

Step 5: First Run

Open a terminal in any project directory and run:

“`bash

claude

“`

Claude Code will start. It will use the MiniMax model (since the environment variables override the defaults). You can verify the model by running:

“`bash

claude -p “what model are you using”

“`

If the model field shows `MiniMax-M3` or whatever you set, you’re connected.

Step 6: Set API Timeout (Optional but Recommended)

If you work with large repos or long prompts, the default timeout can be tight. Add a generous timeout:

“`bash

macOS / Linux

echo ‘export API_TIMEOUT_MS=”3000000″‘ >> ~/.bashrc

source ~/.bashrc

Windows PowerShell

[System.Environment]::SetEnvironmentVariable(“API_TIMEOUT_MS”, “3000000”, “User”)

“`

Common Issues

“Invalid API key”

Double-check that the key was copied correctly and that you’re using the right endpoint. The MiniMax API uses a different URL path than OpenAI’s:

– ❌ `https://api.minimaxi.com/v1/chat/completions` (OpenAI-style — wrong here)

– ✅ `https://api.minimaxi.com/anthropic` (Anthropic-compatible — correct)

“Model not found”

The MiniMax model names are case-sensitive. Common working values:

– `MiniMax-M3` (latest, recommended)

– `MiniMax-M2.7` (previous version, stable)

– `MiniMax-M2.7-highspeed` (faster M2.7 tier)

“Onboarding required” prompt keeps appearing

The `.claude.json` file isn’t being read. Check the location:

– macOS/Linux: `~/.claude.json` (note the dot, no nested folder)

– Windows: `%USERPROFILE%\.claude.json`

The file just needs `{“hasCompletedOnboarding”: true}`.

Slow responses

This is the API endpoint, not your machine. If M3 is too slow, try `MiniMax-M2.7-highspeed` — same model, faster tier.

Daily Workflow

Once set up, your daily workflow is identical to the default Claude Code experience:

“`bash

Start Claude Code in your project

cd ~/projects/my-app

claude

Common commands

/help

/clear

/agents

“`

Slash commands, tool calling, code editing — everything works exactly the same. The only thing that’s different is which model is on the other end.

Cost Comparison

Rough monthly cost for a solo developer using Claude Code ~3 hours/day:

Setup	Approx. monthly
:—	:—
Anthropic default (Sonnet 4.5)	$80-150
MiniMax M3 (current setup)	$5-15
OpenAI GPT-5 mini	$15-30
DeepSeek V4 Pro	$10-20

The savings are real. M3 handles 95% of what Sonnet does, at roughly 1/10th the cost for an individual user.

When to Switch Back

Some tasks still benefit from the actual Claude model:

– Hard reasoning puzzles where M3 makes mistakes

– Tasks requiring extremely careful code review where Anthropic’s models still have a slight edge

– Work where you need the highest-stakes output and cost isn’t the bottleneck

For most daily coding work, M3 is more than good enough.

—

Total time: 10 minutes. After setup, Claude Code is indistinguishable from the default — except much cheaper.

If you hit any snags with the setup, drop me a note. I do this configuration as part of my remote deployment service.

One email per week when I publish a new article or update an existing one. New AI tool reviews, deployment updates, behind-the-scenes notes. No marketing, no spam, unsubscribe in one click.

Subscribe to AimActok Weekly

Or learn more · RSS feed

One email per week when I publish a new article or update an existing one. No marketing, no spam.

→ Subscribe to the newsletter · RSS