Provides NACE Rev. 2.1 codes with browse, lookup, search, and fuzzy matching for MCP clients.
Configuration
View docs{
"mcpServers": {
"dejo90-nace-mcp": {
"command": "npx",
"args": [
"nace-mcp"
]
}
}
}You can access the NACE Rev. 2.1 economic activity codes through a compact MCP server that loads all 1,047 codes into memory at startup and provides four practical tools for browsing, looking up, searching, and fuzzy-matching. It is designed to help AI agents categorize activities quickly and consistently without exposing the entire dataset.
You connect any MCP-compatible client to the NACE MCP server and use its tools to find the right NACE code for a given activity. Use nace_suggest to get the most relevant codes from a free-text description, nace_get to fetch full details for a specific code, nace_browse to explore the hierarchy by parent or top-level sections, and nace_search to find activities by keyword. Each tool returns structured results that you can use to drill down to the exact class you need.
Prerequisites you need before installing: Node.js and npm are required to run the server locally.
Install and run locally by following these steps.
git clone https://github.com/dejo90/nace-mcp.git
cd nace-mcp
npm install
npm run build
node dist/index.jsAdd the MCP server configuration to your client so it can be discovered and used by your AI agent. The server exposes the nace-mcp endpoint via the MCP runtime.
Example MCP client configuration (stdio-based) shown below. This tells the client how to start and where the server runs.
{
"mcpServers": {
"nace-mcp": {
"command": "npx",
"args": ["nace-mcp"]
}
}
}If you plan to inspect or debug the MCP server, you can use the MCP Inspector tooling to connect to the running process and observe the tool interactions.
npx @modelcontextprotocol/inspector node dist/index.jsReturns full details for a single NACE code given a code like A, 01, 25.1, or 25.11.
Returns direct children of a given code, or all top-level sections when no parent is provided. Limits to 50 results.
Performs a case-insensitive substring search across activity labels, returning up to 10 matches.
Fuzzy-matchs a free-text activity description to NACE codes, returning the top 5 candidates with a brief explanation of the match.