OpenCode

OpenCode is an open-source terminal AI coding agent by SST, with multi-provider support and LSP integration.

Option 1: Setup wizard
Option 2: Manual setup

Run:

npx -y mmx-cli@latest agent setup

Select OpenCode. If OpenCode isn’t installed, select it again when prompted to choose which tools to install. The wizard installs the official npm package, verifies its version, and sets a MiniMax default model.Run opencode when setup finishes.

Install OpenCode

curl -fsSL https://opencode.ai/install | bash
# or: npm i -g opencode-ai

Authenticate

Run opencode auth login, then when prompted, search and pick MiniMax Token Plan (minimax.io) (the provider name in OpenCode), and enter your M Plan API key.

Launch

Run opencode and select MiniMax-M3.1-Flash from the model list. If your version does not list it, add the following to ~/.config/opencode/opencode.json and run it again:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "minimax-coding-plan": {
      "models": {
        "MiniMax-M3.1-Flash": {
          "name": "MiniMax-M3.1-Flash",
          "limit": { "context": 524288, "output": 131072 }
        }
      }
    }
  }
}

This example sets a 512K (524288 tokens) client context window to balance cost and usability; the model itself supports up to 1M. See the 512K guidance.Both context and output are required inside limit. Leaving either one out makes OpenCode fail validation and refuse to start. output is the maximum tokens per response; OpenCode may lower it further based on its own ceiling and the room left in the context window, so setting the model’s absolute maximum has no effect.

When configuring through tools such as CC Switch

When using third-party tools such as CC Switch to configure MiniMax in OpenCode, check that the protocol and Base URL match. For the Anthropic-compatible protocol (@ai-sdk/anthropic), use https://api.minimax.io/anthropic/v1 as the Base URL.To adjust the configuration manually, open ~/.config/opencode/opencode.json and add or update these fields:

provider.minimax-coding-plan.npm: set to @ai-sdk/anthropic.
provider.minimax-coding-plan.options.baseURL: set to https://api.minimax.io/anthropic/v1.

The example below shows where these two fields belong and retains the models configuration from the preceding example. Merge these two changes into your existing configuration, keeping your M Plan API key, other models, and settings unchanged:

{
  "provider": {
    "minimax-coding-plan": {
      "npm": "@ai-sdk/anthropic",
      "options": {
        "baseURL": "https://api.minimax.io/anthropic/v1"
      },
      "models": {
        "MiniMax-M3.1-Flash": {
          "name": "MiniMax-M3.1-Flash",
          "limit": { "context": 524288, "output": 131072 }
        }
      }
    }
  }
}

This adapter appends /messages to the Base URL, so keep /anthropic/v1. When changing protocols, also check that the Base URL matches.

Troubleshooting

Web search (web-search) not working?

Some older versions of OpenCode may not enable web search by default, so the model falls back to fetching pages repeatedly or running curl. If this happens, enable web search by setting the environment variables described in the OpenCode documentation.

Overview

Subscription

Guides

Announcements

Troubleshooting