Skip to main content
Route every model call OpenCode makes through the AI Gateway by editing a single configuration file. Requests appear in Traces automatically, and API key usage limits are enforced per workspace.

​
Prerequisites

​
Setup

Edit ~/.config/opencode/config.json and add an Orq.ai provider using the @ai-sdk/openai-compatible package:
Model key must match the AI Gateway model ID exactly. The key in models (e.g. openai/gpt-5.4-mini) is sent verbatim to the AI Gateway as the model field. Using any other alias returns Model not found. Copy the exact model ID from AI Gateway > Models. The name field is display-only in OpenCode’s model picker and has no effect on routing.

​
Smart Router

Point OpenCode at a Smart Router to let the AI Gateway select the best model automatically:
Replace acme@orq/company-approved-models with the router name shown in AI Gateway > Smart Router.

​
Skills

Orq Skills add pre-built agentic workflows to OpenCode for the full Build, Evaluate, Optimize lifecycle.

​
Installation

Set an API key and connect the Orq MCP server first. Skills are then triggered by describing what is needed.

Orq Skills

The full catalogue of skills and slash commands.
Slash commands (/orq:quickstart, /orq:traces, and others) are only available in Claude Code.

​
Troubleshooting

The model key in config.json does not match the model ID in the AI Gateway. Open AI Gateway > Models, copy the exact model ID (e.g. openai/gpt-5.4-mini), and use it as the key in the models object.
Confirm echo $ORQ_API_KEY returns the expected value. The {env:ORQ_API_KEY} placeholder reads the variable at launch time. If the variable is not set, OpenCode sends an empty key.

​
Verification

Send a prompt in OpenCode. The response appears in the editor and the trace appears in Traces with the model identifier and cost breakdown.
Trace detail showing an OpenCode request with model identifier, token counts, latency, and cost breakdown.

An OpenCode request routed through the AI Gateway, visible in Traces.