AI Providers

Cosmo supports multiple AI providers — Anthropic, OpenAI, and Azure AI — with the same interface. You can switch providers at any time without changing your workflow.

Supported Models

Model Tiers

• Fast — Optimized for speed and cost. Good for quick questions, titles, and simple tasks.
• Balanced — Best trade-off between capability and cost. Recommended for most work.
• Premium — Maximum capability. Use for complex reasoning, long documents, and nuanced tasks.

Switching Models

You can switch models at any time using the model selector in the chat header. Click the current model name to see all available models grouped by tier.

When you switch models mid-conversation:

• Your conversation history is preserved
• The context window meter updates to reflect the new model's limit
• If your conversation is close to the new model's context limit, Cosmo may auto-compact older messages

Provider-Specific Capabilities

Some features are only available with certain providers. Cosmo handles this gracefully — if a capability isn't available with your current provider, the core workflow still works, you just won't have that specific feature.

Web Search (Anthropic)

When using Anthropic as your provider, Cosmo can search the web during conversations. This is a server-side capability — Anthropic handles the search execution, and results are incorporated into Cosmo's response automatically.

Web search is enabled by default. No configuration is needed.

Other providers may add web search support in the future — when they do, it will work through the same interface with no changes needed on your end.

How Provider Switching Works

All providers use the same chat interface, tool system, and streaming format. When you switch providers:

• Your conversation history is preserved
• Connected integrations (MCP servers) continue to work
• Artifacts, tasks, and workspaces are unaffected
• Provider-specific capabilities (like web search) are automatically enabled or disabled based on what the new provider supports

Model Availability

Model availability can be controlled at the organization level. Organization admins can configure which models are available to team members. This is useful for managing costs — for example, restricting premium models to specific team members or workspaces.