Setting up MiniMax M2.1 with OpenClaw

I often find myself frustrated when a coding assistant starts wasting tokens on long-winded explanations or struggles with complex multi-language projects. I want a model that gets straight to the point and handles languages like Rust or Go without me having to fix basic syntax errors.

MiniMax M2.1 is built for these real-world tasks. It focuses on concise responses and better tool compatibility, making it a great choice for app development or office-style workflows. I’ve found its ability to handle composite instructions particularly helpful when I need to manage tight constraints.

What You’ll Need

• An active MiniMax Coding Plan (for OAuth) or a MiniMax API Key.
• OpenClaw installed on your machine.
• LM Studio (only if you plan to run it locally).

Quick Start

I recommend the OAuth method if you have a MiniMax Coding Plan. It is the fastest way to get running without managing API keys.

1. Enable the plugin and restart the gateway:

   openclaw plugins enable minimax-portal-auth
   openclaw gateway restart

2. Run the onboarding command:

   openclaw onboard --auth-choice minimax-portal

3. Select your endpoint:

   • Global: International users (api.minimax.io)
   • CN: Users in China (api.minimaxi.com)

Choose Your Setup

If you prefer using an API key or need a fallback configuration, you can use the interactive CLI or edit your config file directly.

Option 1: Using the CLI Wizard

1. Run openclaw configure.
2. Select Model/auth.
3. Choose MiniMax M2.1.
4. Follow the prompts to set your default model.

Option 2: Manual API Configuration

For a hosted setup using the Anthropic-compatible API, add this to your openclaw.json:

{
  env: { MINIMAX_API_KEY: "sk-..." },
  agents: { defaults: { model: { primary: "minimax/MiniMax-M2.1" } } },
  models: {
    mode: "merge",
    providers: {
      minimax: {
        baseUrl: "https://api.minimax.io/anthropic",
        apiKey: "${MINIMAX_API_KEY}",
        api: "anthropic-messages",
        models: [
          {
            id: "MiniMax-M2.1",
            name: "MiniMax M2.1",
            reasoning: false,
            input: ["text"],
            cost: { input: 15, output: 60, cacheRead: 2, cacheWrite: 10 },
            contextWindow: 200000,
            maxTokens: 8192,
          },
        ],
      },
    },
  },
}

Option 3: Fallback Setup

If you want to keep Claude Opus 4.6 as your primary model but have MiniMax M2.1 take over if the first one fails, use this structure:

{
  env: { MINIMAX_API_KEY: "sk-..." },
  agents: {
    defaults: {
      models: {
        "anthropic/claude-opus-4-6": { alias: "opus" },
        "minimax/MiniMax-M2.1": { alias: "minimax" },
      },
      model: {
        primary: "anthropic/claude-opus-4-6",
        fallbacks: ["minimax/MiniMax-M2.1"],
      },
    },
  },
}

Troubleshooting

“Unknown model: minimax/MiniMax-M2.1”

I’ve seen this happen when the MiniMax provider isn’t detected. This usually means the provider entry is missing or the environment key isn’t set. You can fix this by:

• Upgrading to version 2026.1.12 or later and restarting the gateway.
• Running openclaw configure and selecting MiniMax M2.1.
• Manually adding the models.providers.minimax block to your config.

Remember that model IDs are case-sensitive. Use minimax/MiniMax-M2.1 or minimax/MiniMax-M2.1-lightning. You can verify your setup with:

openclaw models list

If you need more help with your specific configuration, check out the AI Setup Assistant.

What’s Next

• Model Provider Concepts
• MiniMax OAuth Plugin Details