MCP Tool Extensions
MCP (Model Context Protocol) is Eidolon's tool extension mechanism. By configuring MCP servers, you can give AI additional capabilities like accessing databases, calling external APIs, and more.
For Advanced Users
MCP configuration requires some technical background. If you're unfamiliar with JSON configuration and command lines, feel free to skip this chapter.
What Is MCP?
MCP is a standardized protocol that enables AI to call external tools. Think of it as installing "plugins" for AI:
- Configure an MCP server = Add a new capability to AI
- AI automatically decides when to call these tools during conversations
- Tool execution results are returned to the conversation
Configuring MCP
- Go to Settings → MCP
- Enter MCP configuration in the JSON editor
- Click Save
- Click Start to run the configured servers
Configuration Format
{
"mcpServers": {
"server-name": {
"command": "path/to/executable",
"args": ["arg1", "arg2"],
"env": {
"ENV_VAR_NAME": "value"
}
}
}
}
Configuration Example
{
"mcpServers": {
"my-search": {
"command": "npx",
"args": ["-y", "@example/mcp-search"],
"env": {
"API_KEY": "your-api-key"
}
}
}
}
Managing MCP Servers
| Action | Description |
|---|---|
| Save Config | Save JSON configuration locally |
| Start Servers | Run all configured MCP servers |
| Stop Servers | Stop running MCP servers |
| View Status | Check each server's running status |
Security Notes
warning
- MCP servers run in your local environment — only configure servers you trust
- Do not expose sensitive information (like API keys) in the configuration
- Environment variable secrets are stored locally only and are not uploaded to the server
Common Issues
MCP server failed to start?
- Check that the
commandpath is correct - Confirm
argsparameters are correct - Verify environment variables are set completely
AI not calling MCP tools?
- Confirm the server is running
- Try explicitly prompting AI to use a specific tool in conversation
- Check that tool descriptions are clear