OpenCode

OpenCode is an open-source AI coding agent that runs in the terminal and can be used alongside any editor or IDE.

Key Features

• LSP support that loads the right language tooling for the model
• Multi-session workflows so you can run multiple agents on the same project
• Shareable session links for collaboration and debugging
• MCP support for external tools and data sources
• Editor-agnostic workflow that works in any IDE

Installation and Setup

Proxy usage

Like most command-line tools, OpenCode may require proxy settings in your shell environment. If you are on Windows and need a proxy, set these variables before starting OpenCode:

http_proxy=<your-proxy-url:port>
https_proxy=<your-proxy-url:port>

1. Follow the OpenCode installation instructions on AI Coding Plugins.

2. Run OpenCode:

   opencode

3. In OpenCode, run the following command to choose your provider and sign in with your API key:

   /connect

   Select AI Foundation Services from the provider list, then provide your API key when prompted. You can generate an API key in LLM Serving API Keys.

You can now start a chat and use OpenCode.

Example Config

If you’d rather configure OpenCode declaratively (or pin a default model and base URL), create opencode.json at your project root or ~/.config/opencode/opencode.json for a global default:

{
  "$schema": "https://opencode.ai/config.json",
  "model": "openai/gpt-4.1", // Set your default model (format: provider_id/model_id)

  "provider": {
    "openai": {
      "options": {
        "baseURL": "https://llm-server.llmhub.t-systems.net/v2"
      }
    }
  },

  // Set permissions for built-in tools
  "permission": {
    "edit": "allow",
    "websearch": "allow"
  }
}

OpenCode automatically picks up OPENAI_API_KEY from your environment, so the config only needs the baseURL. If you set the variable during the Quickstart, no further key configuration is needed here.