Augment Code for OpenCode

Use Claude and GPT models at zero cost in OpenCode through Augment Code's infrastructure.

Quick Start

1. Install Auggie CLI

npm install -g @augmentcode/auggie

2. Run the setup script

git clone https://github.com/hletrd/augment-opencode.git
cd augment-opencode
./setup.sh

Or configure manually — add to ~/.config/opencode/opencode.json:

{
  "plugin": ["opencode-augment-auth"],
  "model": "augment/claude-opus-4-6"
}

3. Authenticate in OpenCode

opencode auth login

You'll see two connection options:

Option A: Auggie CLI Session (recommended)

1. Run auggie login — this opens a browser to authenticate with Augment Code and saves credentials to ~/.augment/session.json
2. In OpenCode, select "Auggie CLI Session (auggie login)"
3. The plugin auto-detects the session file — no further input needed

Option B: API Key

1. Run auggie tokens print to get your API token
2. Run auggie tokens print --api-url to get the API URL
3. In OpenCode, select "API Key (from auggie tokens print)"
4. Enter the token and URL when prompted

4. Use it

opencode

That's it. The plugin handles everything automatically:

• Registers the augment provider with all 8 models
• Starts an embedded HTTP server on a random port (no conflicts between instances)
• Routes requests through the Auggie SDK to Augment Code's API
• Self-heals if the embedded server crashes

Available Models

Use augment/<model-id> in OpenCode (e.g., augment/claude-opus-4-6).

Model ID	Description	Context	Output
claude-opus-4-6	Most capable Claude	200K	32K
claude-opus-4-5	Previous Opus gen	200K	32K
claude-sonnet-4-5	Balanced performance	200K	16K
claude-sonnet-4	Previous Sonnet gen	200K	16K
claude-haiku-4-5	Fast, lightweight	200K	8K
gpt-5	GPT-5	128K	16K
gpt-5-1	Strong reasoning	128K	16K
gpt-5-2	Smarter, slower	128K	16K

How It Works

OpenCode → Plugin (config + auth hooks) → Embedded Server → Auggie SDK → Augment Code API → Claude/GPT

1. The config hook auto-injects the augment provider with all models into OpenCode
2. The auth hook validates credentials (session file or API key) and starts an embedded HTTP server
3. The embedded server wraps the Auggie SDK with an OpenAI-compatible API
4. Each OpenCode instance gets its own server on an OS-assigned port (no collisions)
5. Streaming uses Server-Sent Events with agent_message_chunk → delta.content and agent_thought_chunk → delta.reasoning_content

Troubleshooting

Issue	Solution
"Session file not found"	Run auggie login to create ~/.augment/session.json
Auth login fails	Run auggie login again, or try the API Key method instead
"Token is required"	Run auggie tokens print and paste the output
Model not found	Use hyphenated IDs: claude-opus-4-6, not claude-opus-4.6
Switch auth method	Run opencode auth login again and pick the other option
Verbose logging	Set AUGMENT_DEBUG=true before running OpenCode

Advanced: Standalone Server

If you want to use Augment Code models with other OpenAI-compatible clients (not just OpenCode), you can run the standalone HTTP server directly:

cd augment-opencode
npm install
npm start

The server starts on http://localhost:8765.

API Endpoints

Endpoint	Method	Description
/v1/chat/completions	POST	Chat completions (streaming supported)
/v1/models	GET	List available models
/v1/models/{id}	GET	Get model details
/health	GET	Detailed health check with metrics
/version	GET	Server version and config
/metrics	GET	Raw request metrics

Example

curl http://localhost:8765/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "claude-haiku-4-5", "messages": [{"role": "user", "content": "Hello!"}]}'

Environment Variables

Variable	Default	Description
PORT	8765	Server port
DEBUG	false	Enable debug logging
REQUEST_TIMEOUT_MS	300000	Request timeout (5 min)
SHUTDOWN_TIMEOUT_MS	30000	Graceful shutdown timeout (30s)

Project Structure

augment-opencode/
├── plugin/                  # OpenCode plugin
│   ├── src/
│   │   ├── plugin.ts        # Config + auth hooks, embedded server lifecycle
│   │   ├── server.ts        # Embedded HTTP server wrapping Auggie SDK
│   │   └── models.ts        # Shared model definitions
│   └── package.json
├── src/
│   └── server.ts            # Standalone HTTP server (advanced)
├── setup.sh                 # Automated setup script
└── README.md

License

MIT