OpenCode

OpenCode can use models from a self-hosted (OpenAI-compatible) provider. To enable this, set the LOCAL_ENDPOINT environment variable to the URL of your local endpoint. This tells OpenCode to route model requests to your specified address instead of the default provider.

LOCAL_ENDPOINT=http://localhost:1235/v1

For scripting or automation, you can run OpenCode without the TUI by passing a prompt directly via the --prompt (or -p) flag. In this mode, all permissions are auto-approved, and the output is printed to standard output before the application exits.

Output Formats

Use the --output-format (or -f) flag to specify the output type:

• text: Plain text output (default)
• json: Output wrapped in a JSON object

Usage Examples

# Run a single prompt and print the AI's response to the terminal
opencode -p "Explain the use of context in Go"

# Get response in JSON format
opencode -p "Explain the use of context in Go" -f json

# Run without showing the spinner (useful for scripts)
opencode -p "Explain the use of context in Go" -q

Once you have generated opencode-schema.json, you can use it in the following ways:

1. Editor Integration: Configure editors like VS Code or JetBrains IDEs to use the generated schema for .opencode.json files to get real-time validation and autocompletion.
2. Validation Tools: Use CLI tools like jsonschema to programmatically validate your configuration files against the schema.
3. Documentation: Refer to the schema to understand available configuration options, types, and default values.

To build OpenCode manually, ensure you have Go 1.24.0 or higher installed. Clone the repository, use go build to generate the binary, and then execute the resulting file.

# Clone the repository
git clone https://github.com/opencode-ai/opencode.git
cd opencode

# Build
go build -o opencode

# Run
./opencode

You can install OpenCode using several methods depending on your operating system and preferred package manager.

Using the Install Script

To install the latest version or a specific version via curl:

# Install the latest version
curl -fsSL https://raw.githubusercontent.com/opencode-ai/opencode/refs/heads/main/install | bash

# Install a specific version
curl -fsSL https://raw.githubusercontent.com/opencode-ai/opencode/refs/heads/main/install | VERSION=0.1.0 bash

Using Homebrew (macOS and Linux)

brew install opencode-ai/tap/opencode

Using AUR (Arch Linux)

# Using yay
yay -S opencode-ai-bin

# Using paru
paru -S opencode-ai-bin

Using Go

go install github.com/opencode-ai/opencode@latest

Use the cmd/schema tool to generate a JSON Schema file. This schema is used to validate .opencode.json configuration files and enable autocompletion in supported editors.

Run the following command to output the schema to a file named opencode-schema.json:

go run cmd/schema/main.go > opencode-schema.json

Copilot support is experimental. To use it, you must have Copilot chat enabled in your GitHub settings and use one of the following:

• VSCode GitHub Copilot chat extension
• GitHub gh CLI
• Neovim GitHub Copilot plugin (copilot.vim or copilot.lua)
• A GitHub token with copilot permissions

Ensure you have authenticated the tool with your GitHub account. If using an explicit token, set the $GITHUB_TOKEN environment variable or add it to opencode.json under providers.copilot.apiKey.

OpenCode is a terminal-based AI assistant for software development. It can be used in two primary modes:

1. Interactive Mode: Launches a full Terminal User Interface (TUI) with an interactive chat interface, code analysis, and LSP integration.
2. Non-Interactive Mode: Allows you to send a single prompt and receive a response directly in your terminal, which is useful for scripting or quick queries.

To start the interactive TUI, simply run the command without additional arguments.

# Run in interactive mode
opencode

OpenCode integrates with LSP to provide code intelligence. Currently, the AI assistant uses LSP primarily through the diagnostics tool to check for errors and suggest fixes.

Configure language servers in the lsp section of your configuration file.

{
  "lsp": {
    "go": {
      "disabled": false,
      "command": "gopls"
    },
    "typescript": {
      "disabled": false,
      "command": "typescript-language-server",
      "args": ["--stdio"]
    }
  }
}

To specify a particular model from your self-hosted provider, update the agents section in your OpenCode configuration file. You can define the model name and set the reasoningEffort level for specific agents (e.g., the coder agent).

{
  "agents": {
    "coder": {
      "model": "local.granite-3.3-2b-instruct@q8_0",
      "reasoningEffort": "high"
    }
  }
}