Conductor Integration

Conductor runs coding agents in separate workspaces. Its bundled OpenCode agent includes an LLM Gateway provider, so you can connect from Settings without installing a separate CLI or entering a base URL.

Verified with Conductor 0.87.4.

Prerequisites

• Install Conductor from the official site.
• Create an LLM Gateway API key, or use a DevPass plan key.
• Open a repository and workspace in Conductor.

Connect LLM Gateway

1. Open Settings → Agents → OpenCode. Some versions label Agents as Harnesses.
2. Under Providers, click Configure.

3. If LLM Gateway is missing, click View all providers, then search for LLM Gateway.

4. Expand LLM Gateway, paste your API key into LLMGATEWAY_API_KEY, and click Save. A checkmark indicates that a key is configured.

The built-in provider supplies the gateway endpoint. Use your pay-as-you-go key for usage-based billing or your DevPass key for your coding plan.

Choose a model

1. Close the provider dialog and click Configure under Models.
2. Search for a model and check its entry under llmgateway. Choose a model with tool calling from the live catalogue.

For DevPass, choose a model included in your plan under llmgateway. The llmgateway-providers entries pin an upstream provider and are not supported on coding plans. The model shown in the screenshot is an example; availability changes with the catalogue and your account.

3. Open Settings → Default models and add your model from the OpenCode column to your Loadout. To change an existing slot, open its menu, choose Agent → OpenCode, then select your model under Model.

4. Return to your workspace, open a new chat, and choose that model from the agent picker below the message box. The picker's Edit button takes you back to Default models.

Verify the connection

Send a small prompt first:

Reply with exactly: Hello from LLM Gateway! Do not use tools or read or change any files.

In your LLM Gateway dashboard, select the project that issued the key and open Activity from the sidebar to inspect the request. DevPass users can review usage under Coding Agents; Conductor runs these requests through OpenCode.

Once connected, ask the agent to make a small change and run the relevant tests. Review the tool output and workspace diff before keeping the change.

Troubleshooting

LLM Gateway is missing: click View all providers and search again. If it still does not appear, update Conductor.

The model is missing from the chat picker: enable it under Agents → OpenCode → Models, then add it to Default models → Loadout.

Authentication fails: check that the provider key was saved and is active in the matching dashboard.

Requests use another provider: select the model under llmgateway, then confirm that the chat is using your OpenCode loadout entry.

For executable overrides and other provider configuration options, see Conductor's OpenCode documentation.