Centralized API Configuration & Model Assignment for Custom Modes

[kfuglsang]

Summary
Enable organizations to centrally define API configurations (including model selection) and reference them in Custom Modes, so that model usage can be standardized and optimized without requiring each user to configure Roo Code individually.

Problem
Currently, Roo Code’s Sticky Models remember the last model a user chose for a mode, but this is per-user and manual. This:

• Creates inconsistency in results when teams share modes.
• Wastes time for each user to configure models.
• Makes it difficult to enforce cost/performance choices across an organization.

Proposed Solution

• Allow named API configs (e.g., fast-gpt4o, cost-optimized-gpt-4-mini) to be defined in a shared config file or organizational settings.
• Enable Custom Modes to reference these API configs instead of a fixed local model.
• Optionally lock configurations or allow overrides.
• Maintain backward compatibility with current Sticky Model behavior if no shared config is set.

Benefits

• Consistency – All team members use the same models for the same modes.
• Cost control – Change models centrally to manage spend.
• Faster onboarding – New users get correct settings automatically.
• Governance – Central oversight of AI model usage.

Example
Our team uses Roo Code for developers, architects, and sales with multiple Custom Modes. Each mode should always use the same model for everyone, but today each person has to manually pick it. Central configs would solve this and allow us to optimize cost and performance centrally.