Models
Choose the agent's Fast and Pro models and configure workspace provider credentials for BYOK usage.
Model settings control which AI models the agent uses. You can choose separate models for faster work and heavier reasoning, and optionally provide workspace credentials for supported providers.
Where to find it
Open Settings → Models.
Choose Fast and Pro models
The agent has two model slots:
| Slot | Typical use | Catalog options |
|---|---|---|
| Fast model | Quicker, lighter work | GPT-5.6 Luna from OpenAI and Claude Sonnet 5 from Anthropic |
| Pro model | Heavier reasoning | GPT-5.6 Sol from OpenAI and Claude Opus 4.8 from Anthropic |
Models that are eligible for either tier can also appear in the applicable selectors.
The billing hint beneath each selector shows that usage may draw from Endstack credits or from your workspace's OpenAI, Anthropic, or Bedrock credentials when bring-your-own-key (BYOK) is configured for the selected provider.
Configure workspace credentials
The same Settings section lets you manage provider credentials for the workspace:
| Provider | Required values | Available actions |
|---|---|---|
| OpenAI | API key beginning with sk-… | Add, Replace, or Remove |
| Anthropic | API key beginning with sk-ant-… | Add, Replace, or Remove |
| AWS Bedrock | Access key, secret key, and region | Configure the workspace's Bedrock credentials |
A paid seat is still required when BYOK is enabled. When the runtime uses a model from a provider with matching workspace credentials, that usage can skip Endstack AI credits.
How model settings affect the agent
The agent runtime chooses the Fast or Pro slot based on the run. Updating either selector changes the model available for that slot; it does not change which tools the agent can use or whether those tools require approval.
Configure tool permissions separately under Autonomy and approvals.