MCP Server Configuration

RAPID supports MCP (Model Context Protocol) servers to extend AI agent capabilities with external tools, data sources, and services.

Overview

MCP servers provide AI agents with access to:

• Documentation context - Library and framework documentation
• Web search - Real-time information retrieval
• Browser automation - Web scraping and testing
• Database access - Query and manage databases
• File system - Read and write files
• Memory - Persistent knowledge graphs

Quick Start

Initialize with MCP Servers

When initializing a new RAPID project, you can specify which MCP servers to enable:

# Default setup (context7 + tavily)
rapid init

# Specify servers
rapid init --mcp context7,tavily,playwright

# Skip MCP configuration
rapid init --no-mcp

Managing MCP Servers

Use the rapid mcp command to manage servers:

# List configured servers
rapid mcp list

# Show available templates
rapid mcp list --templates

# Add a server from template
rapid mcp add playwright

# Add a custom server
rapid mcp add myserver --type remote --url https://api.example.com/mcp

# Enable/disable servers
rapid mcp disable tavily
rapid mcp enable tavily

# Remove a server
rapid mcp remove playwright

# Show server status
rapid mcp status

# Regenerate config files
rapid mcp sync

Available Server Templates

RAPID includes built-in templates for popular MCP servers:

Server	Type	Description	Required Secrets
context7	Remote	Documentation context for libraries	CONTEXT7_API_KEY
tavily	Remote	Web search and data extraction	TAVILY_API_KEY
playwright	Stdio	Browser automation and web scraping	None
github	Stdio	GitHub operations (PRs, issues)	GITHUB_TOKEN
filesystem	Stdio	File system access	None
memory	Stdio	Persistent knowledge graph	None
postgres	Stdio	PostgreSQL database access	DATABASE_URL
slack	Stdio	Slack messaging	SLACK_TOKEN
fetch	Stdio	HTTP content retrieval	None
sqlite	Stdio	SQLite database access	None

Configuration

In rapid.json

MCP servers are configured in the mcp section:

{
  "mcp": {
    "configFile": ".mcp.json",
    "servers": {
      "context7": {
        "enabled": true,
        "type": "remote",
        "url": "https://mcp.context7.com/mcp",
        "headers": {
          "Context7-API-Key": "${CONTEXT7_API_KEY}"
        }
      },
      "tavily": {
        "enabled": true,
        "type": "remote",
        "url": "https://mcp.tavily.com/mcp",
        "headers": {
          "Authorization": "Bearer ${TAVILY_API_KEY}"
        }
      },
      "playwright": {
        "enabled": true,
        "type": "stdio",
        "command": "npx",
        "args": ["@playwright/mcp@latest"]
      }
    }
  }
}

Server Types

Remote (HTTP) Servers

Remote servers communicate over HTTP:

{
  "myserver": {
    "type": "remote",
    "url": "https://api.example.com/mcp",
    "headers": {
      "Authorization": "Bearer ${API_KEY}"
    }
  }
}

Stdio Servers

Stdio servers run as local processes:

{
  "myserver": {
    "type": "stdio",
    "command": "npx",
    "args": ["@example/mcp-server"],
    "env": {
      "API_KEY": "${API_KEY}"
    }
  }
}

Variable Substitution

Server configurations support variable substitution:

Syntax	Description
${VAR}	Environment variable
${env:VAR}	Environment variable from container
${localEnv:VAR}	Environment variable from host

Generated Config Files

RAPID generates two config files from your rapid.json:

.mcp.json

Used by Claude Code and other MCP-compatible tools:

{
  "mcpServers": {
    "context7": {
      "type": "http",
      "url": "https://mcp.context7.com/mcp",
      "headers": {
        "Context7-API-Key": "${CONTEXT7_API_KEY}"
      }
    }
  }
}

opencode.json

Used by OpenCode with its specific format:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "context7": {
      "type": "remote",
      "url": "https://mcp.context7.com/mcp",
      "headers": {
        "Context7-API-Key": "{env:CONTEXT7_API_KEY}"
      }
    }
  }
}

Secrets for MCP Servers

Most MCP servers require API keys or tokens. Configure these in the secrets.items section:

{
  "secrets": {
    "provider": "1password",
    "vault": "Development",
    "items": {
      "CONTEXT7_API_KEY": "op://Development/Context7/api-key",
      "TAVILY_API_KEY": "op://Development/Tavily/api-key",
      "GITHUB_TOKEN": "op://Development/GitHub/pat"
    }
  }
}

Getting API Keys

Server	Where to Get Key
Context7	context7.com
Tavily	tavily.com
GitHub	github.com/settings/tokens
Slack	api.slack.com/apps

Using MCP Servers

Once configured, MCP servers are automatically available to your AI agents during rapid dev:

1. RAPID loads secrets from your secret provider
2. Generates .mcp.json and opencode.json with resolved references
3. Sets MCP_CONFIG_FILE environment variable
4. Launches the agent with MCP servers available

Example Workflow

# 1. Initialize project with MCP
rapid init --mcp context7,tavily

# 2. Add your API keys to secrets
# Edit rapid.json secrets.items

# 3. Verify secrets are accessible
rapid secrets verify

# 4. Start development with MCP servers
rapid dev

Adding Custom MCP Servers

You can add any MCP-compatible server:

# Add a custom remote server
rapid mcp add myapi \
  --type remote \
  --url https://api.example.com/mcp \
  --header "Authorization=Bearer ${API_KEY}"

# Add a custom stdio server
rapid mcp add mytool \
  --type stdio \
  --command npx \
  --args "@example/mcp-server,--config,./config.json"

Troubleshooting

Server Not Appearing

If a server doesn’t appear in agent tools:

1. Check it’s enabled: rapid mcp status
2. Verify secrets: rapid secrets verify
3. Regenerate configs: rapid mcp sync

Authentication Errors

If you get authentication errors:

1. Verify the API key is correct
2. Check the header format matches the server’s requirements
3. Ensure secrets are loaded: rapid secrets list

Stdio Server Fails to Start

If a stdio server won’t start:

1. Check the command is installed: which npx
2. Verify the package exists
3. Check for error messages in agent output