MCP integration
The Model Context Protocol (MCP) is an open protocol that enables large language models to interact with applications through standardized interfaces.
Chef-cli provides MCP integration through the chef-cli mcp initial-mapping
subcommand.
This integration allows AI assistants to programmatically edit and test initial domain mapping files.
Limitations
MCP integration is experimental. Functionality may vary across different AI models and MCP clients.
Testing shows good results with Cursor and Claude Code. The AI can prepare initial mappings with significant autonomy, though developer oversight remains essential.
Prerequisites
- Chef-cli installed and configured
- MCP-compatible client (Cursor, Claude Code, etc.)
- External domain metadata file
- Empty or partially populated initial domain mapping file
Configuration
Create a .mcp.json
file with the following content:
Cursor setup
- Provide
.cursor/mcp.json
in your project root. - Using the editor Agent mode point the AI to the initial domain mapping file.
- Use the editor to guide the AI to create the initial domain mappings.
- Use
initial-mapping check
to test the initial domain mappings against the local domain metadata file.
Claude Code setup
- Provide
.mcp.json
in your project root. - Using the editor point the AI to the initial domain mapping file.
- Use the editor to guide the AI to create the initial domain mappings.
- Use
initial-mapping check
to test the initial domain mappings against the local domain metadata file.
Other IDEs
You can use the same .mcp.json
file in other IDEs that support MCP.
Usage guidelines
To guide the AI tool, provide the following instructions to your AI assistant:
Adjust file paths according to your project structure.
Available operations
The MCP server provides tools for:
- Testing mappings against metadata
- Adding and removing record type mappings
- Field mapping and unmapping operations
- Validation of mapping file structure
- Testing initial domain mapping files against local metadata files