Registrar un Servidor MCP

MCP (Model Context Protocol) servers extend your AI assistant with external tools. RaiSE manages server registration, configuration, and health checking through the rai mcp command group.

Two Registration Methods

Method 1: Install a Package

For published MCP servers, use rai mcp install:

# npx package (Node.js)
rai mcp install "@upstash/context7-mcp" --type npx --name context7

# uvx package (Python)
rai mcp install "semgrep-mcp" --type uvx --name semgrep --env SEMGREP_APP_TOKEN

# pip package (Python module)
rai mcp install "mcp-server-fetch" --type pip --name fetch --module mcp_server_fetch

Method 2: Scaffold Manually

For custom or local servers, use rai mcp scaffold:

rai mcp scaffold my-server \
  --command "node" \
  --args "path/to/server.js" \
  --env "MY_API_KEY"

Config File Format

Both methods create a YAML config in .raise/mcp/. The catalog format:

# .raise/mcp/catalog.yaml
servers:
  context7:
    package: "@upstash/context7-mcp"
    type: npx
    description: "Documentation lookups for any library via Context7"
    recommended_for: all

  github:
    package: "@modelcontextprotocol/server-github"
    type: npx
    env: [GITHUB_TOKEN]
    description: "GitHub repository operations"
    recommended_for: all

  semgrep:
    package: "semgrep-mcp"
    type: uvx
    env: [SEMGREP_APP_TOKEN]
    description: "SAST + SCA + secret detection"
    recommended_for: all

Config fields: - package — Package identifier for installation - type — npx, uvx, or pip - module — Python module name (pip only) - env — Required environment variable names - description — Human-readable purpose - recommended_for — Language list, or all for universal

Verification

After registration, verify your server works:

# List registered servers
rai mcp list

# Check server health
rai mcp health context7

# List available tools on a server
rai mcp tools context7

# Call a specific tool
rai mcp call context7 resolve-library-id --args '{"libraryName": "react"}'

Example: Registering Context7

Context7 provides documentation lookups for any library. Here is the full flow:

# Install
rai mcp install "@upstash/context7-mcp" --type npx --name context7

# Verify
rai mcp health context7
rai mcp tools context7

No environment variables needed. The server is immediately available to your AI assistant for documentation queries.