Model Providers
Control which AI models your organization can access. This is one of the most important admin controls—it directly affects what's possible in your organization.
Aisle supports three major providers:
- Anthropic - Claude models
- OpenAI - GPT models
- OpenRouter - Access to multiple model providers through one integration

Enabling and Disabling Models
Click into any provider to see all available models. You can enable or disable specific models.
Why you'd disable models:
- Cost control - Some models are expensive. If your team doesn't need GPT-4o for most tasks, disable it and point them to cheaper alternatives.
- Consistency - Too many options creates decision paralysis. Enable the 3-4 models that actually make sense for your team's work.
- Governance - Some models might not meet your compliance requirements. Disable them to prevent accidental use.
Why you'd enable more models:
- Different strengths - Some models excel at creative tasks, others at structured data analysis. Different teams might need different models.
- Testing - Enable new models so your team can test them in playgrounds before committing.
- Redundancy - If one provider has an outage, your team can switch to an alternative.
Changes take effect immediately across your entire organization. If you disable a model, it disappears from the chat launcher for all users.
Adding Your Own API Keys
For each provider, you can add your own API access key.
Why use your own keys:
- Cost tracking - API usage shows up on your own provider dashboard
- Rate limits - Your own keys may have higher rate limits than Aisle's shared infrastructure
- Compliance - Some organizations require API calls to go through their own accounts for auditing
To add a key:
- Click into the provider settings
- Find your API key from the provider (OpenAI dashboard, Anthropic console, etc.)
- Paste it into Aisle
- Save
It activates immediately. All API calls for that provider now use your key instead of Aisle's.
Common Scenarios
Team Wants to Test a New Model
- Enable that model in Model Providers
- Let them test in playgrounds
- If it proves valuable, keep it enabled
- If not, disable it to reduce clutter
Cost Control Needed
- Review which models are most expensive
- Disable or restrict access to those models
- Point teams to cheaper alternatives
- Add your own API keys to track usage directly