OpenCode
Configure OpenCode to use BullSequana AI from terminal-first development workflows.
OpenCode connects to BullSequana AI through the BSQAI API.
The BSQAI API sends each request to the server that runs the model, so the stable BullSequana API endpoint is the recommended developer integration path.
What to use
Use:
- API type: OpenAI (Responses API)
- base URL:
https://api.<platform-domain>/v1 - token: JWT or
sk-bsq-...API key
Do not configure OpenCode against an internal model endpoint, such as KServe's.
1. Install OpenCode
Install OpenCode using your preferred installation method.
Examples:
npm install -g opencode-aicurl -fsSL https://opencode.ai/install | bash2. Prepare authentication
Use either:
- a JWT bearer token
- a BullSequana AI API key
For most persistent local tool configurations, API keys are usually more practical. Give the key chat.message.create to generate answers and tenant.read to list models in step 5. If you belong to a group in the tenant, also choose the Team the key's usage counts for. It cannot be changed later, and that team's monthly BCU cap applies to the key. See Create and use an API key.
The configuration below reads the key from the environment rather than storing it, so export it first:
export BSQ_API_KEY="sk-bsq-v1-..."OpenCode sends the key as Authorization: Bearer, which the BSQAI API accepts.
3. Configure the provider
Create or edit ~/.config/opencode/opencode.json.
Example configuration:
{
"$schema": "https://opencode.ai/config.json",
"model": "bsqai/<model-name-from-/v1/models>",
"provider": {
"bsqai": {
"name": "BullSequana AI",
"npm": "@ai-sdk/openai",
"options": {
"baseURL": "https://api.<platform-domain>/v1",
"apiKey": "{env:BSQ_API_KEY}"
},
"models": {
"<model-name-from-/v1/models>": {
"name": "<model-name-from-/v1/models>",
"limit": { "context": <context-window>, "output": <max-output-tokens> }
}
}
}
}
}The "npm": "@ai-sdk/openai" line makes OpenCode use the official OpenAI SDK provider, which talks to the platform's /v1/responses endpoint (the Responses API). Without it, OpenCode uses a generic OpenAI-compatible provider that calls /v1/chat/completions. The BSQAI API serves both.
The limit block gives OpenCode the model's context window and maximum output tokens, which it cannot read from the API. OpenCode measures the session against context and compacts the conversation before it reaches it. Without the block, the context indicator stays at 0% and a long session fails once it reaches the model's limit. Use the values the deployment was started with; the Portal fills them in when the model publishes them.
4. Run OpenCode
Example:
opencodeThe model field above makes this the default. To run one session against a different model, name it as provider/model:
opencode --model bsqai/<model-name-from-/v1/models>5. Check model availability first
Available models are also visible in the Portal.
If you want the API-level source of truth for the current environment, use:
curl "https://api.<platform-domain>/v1/models" \
-H "Authorization: Bearer <your-token>"Why this is the recommended path
This gives OpenCode users:
- a stable API contract
- consistent authentication
- environment portability across clusters
- less coupling to internal platform implementation details
The Portal generates this whole configuration with the model id and base URL already filled in. See Studio — Use model.