Agent Integration

Configure AI providers, select models, and manage agent permissions.

Supported Providers

Provider	Auth method	Notes
Anthropic	API key	Claude models
OpenAI	API key	GPT models
Google	API key / OAuth	Gemini models
GitHub Copilot	OAuth	Requires GitHub Copilot subscription
Groq	API key	Fast inference
DeepSeek	API key	Cost-effective
Mistral	API key	Open-weight models
xAI	API key	Grok models
OpenRouter	API key	Multi-provider gateway
Amazon Bedrock	API key	AWS-hosted models
OpenCode	None	Free tier models

---

Free Models

These models work without any API key:

Model	Reasoning	Tool calling
Minimix	No	Yes
GLM5	No	Yes
KimiK2.5	Yes	Yes

---

Setting Up a Provider

API Key

1. Open Settings

   Navigate to Settings in the sidebar.

2. Select provider

   Choose your provider from the list.

3. Enter API key

   Paste your API key. It is stored locally via OpenCode’s configuration.

Via API:

curl -X POST http://localhost:3001/api/opencode/auth \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <token>" \
  -d '{"providerId": "anthropic", "apiKey": "sk-ant-..."}'

OAuth

Some providers (like GitHub Copilot) support OAuth:

1. Start OAuth

   Click the provider’s Connect button in Settings. This opens the provider’s authorization page.

2. Authorize

   Grant OpenLinear access to the provider.

3. Callback

   The OAuth callback completes automatically and the provider appears as connected.

---

Model Selection

Select which model to use for execution:

1. Open the Model Selector dropdown in the project config panel (top of the kanban board)
2. Browse connected providers and their available models
3. Select a model — it applies to all subsequent executions

Via API:

curl -X POST http://localhost:3001/api/opencode/config/model \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <token>" \
  -d '{"model": "anthropic/claude-sonnet-4-20250514"}'

Get current model:

curl http://localhost:3001/api/opencode/config \
  -H "Authorization: Bearer <token>"

---

Permission System

During execution, the AI agent may request permission to use certain tools (e.g., executing shell commands, writing files). These appear as permission requests in the task detail view.

Permission Responses

Response	Effect
once	Allow this specific action once
always	Always allow this type of action
reject	Deny the action

Responding to Permissions

In the UI, permission requests appear as actionable cards in the task detail view. Via API:

curl -X POST http://localhost:3001/api/tasks/:id/permissions/:permissionId/respond \
  -H "Content-Type: application/json" \
  -d '{"response": "once"}'

---

Provider Status

Check which providers are connected:

curl http://localhost:3001/api/opencode/setup-status \
  -H "Authorization: Bearer <token>"

Returns:

{
  "providers": [
    {"id": "anthropic", "name": "Anthropic", "authenticated": true},
    {"id": "openai", "name": "OpenAI", "authenticated": false},
    {"id": "opencode", "name": "OpenCode", "authenticated": true}
  ],
  "ready": true
}

The ready field is true when at least one provider is authenticated.