MCP Servers & LLMs
Connect external tools and choose AI models to power your knowledge base.
MCP Servers
MCP (Model Context Protocol) servers let you connect external tools and services. This expands what your agents can do.
Adding an MCP Server
- Go to Settings → MCP Servers
- Click “Add Server”
- Enter the server name and URL
- Choose authentication: none, static headers (API key / bearer token), or OAuth for protected remote servers
- Enable the server
Once connected, tools from the MCP server become available to your agents.
LLM Providers
You can choose different AI models for different workspaces:
Self-Hosted Models
Use local models with Ollama for:
- Complete privacy (data never leaves your infrastructure)
- Offline operation
- No usage costs
Cloud Models
Connect to cloud providers like AWS Bedrock for:
- More powerful models when needed
- Research and non-sensitive data
- Additional processing capacity
Choosing a Provider
- Use local models for private or sensitive data
- Use cloud models for research or when you need more processing power
- Mix both: use local for private workspaces, cloud for research workspaces
Workspace-Level Settings
Each workspace can have its own LLM provider. This lets you:
- Keep sensitive data on local models
- Use cloud models for research
- Match the model to the workspace’s needs