- Get Started

Extending: Models

Models are the reasoning engine behind each agent. Extending OpenFlows with a new model is a configuration-only change: you enable the model in the platform's AI catalog, and the team can then route any role to it. No system code is involved.

Where models live

Models are configured in the platform's AI settings (the Coder dashboard), where you enable the models you want available. OpenFlows then references those models for each role — so the catalog lives in the platform, and the per-role assignment lives in your team's registry.

Enabling a new model

  1. Open the platform's AI settings in the Coder dashboard.
  2. Enable the model you want to add to the catalog.
  3. Save. The model is now available for any role to reference.

This is a platform-level change. From here, assigning a role to the new model is a separate step that lives in the agent registry — see Model Routing.

What you should see

Enabling vs. routing

Extending: Models is about making a model available — enabling it in the catalog. Model Routing is about deciding which available model each role uses. Enable first, then route: a model that isn't enabled can't be assigned.

Related extensions

Skills, MCP servers, and models are the three configuration-only capability extensions. They compose: a role can carry several skills, reach several tools, and run on whatever enabled model suits its job.

Where to go next