MCP Server Configuration

CoreLayer connects to external tools and data sources through MCP servers. Each server exposes tools that Jarvis can call during conversations, subject to the Permission Guard.

Server Configuration

Each MCP server is defined by a name, command, and optional arguments:

{
  "mcp": {
    "servers": [
      {
        "name": "filesystem",
        "command": "npx",
        "args": ["-y", "@modelcontextprotocol/server-filesystem", "/workspace"],
        "env": {}
      },
      {
        "name": "github",
        "command": "npx",
        "args": ["-y", "@modelcontextprotocol/server-github"],
        "env": {
          "GITHUB_TOKEN": "${GITHUB_TOKEN}"
        }
      }
    ]
  }
}

Server Types

Stdio Servers

Most MCP servers use stdio (standard input/output) for communication:

{
  "name": "my-server",
  "command": "node",
  "args": ["./my-mcp-server.js"]
}

SSE Servers

For remote servers using Server-Sent Events:

{
  "name": "remote-server",
  "url": "https://my-server.example.com/mcp"
}

Managing Servers

Via Control Center

• Add — Settings → MCP → Add Server
• Remove — click the delete button on a server card
• Restart — click the restart button to reconnect
• View tools — see all tools exposed by a connected server

Connection Status

Tool Discovery

When an MCP server connects, CoreLayer automatically:

1. Discovers all tools exposed by the server
2. Registers them in the Tool Registry
3. Makes them available to Jarvis for tool calling

Tools from MCP servers are treated the same as native tools — they go through the Permission Guard before execution.

Next Steps

• First MCP Server — basic setup guide
• Tool Registry — how tools are managed
• Connect an MCP App — step-by-step walkthrough