Model Context Protocol server for NocoDB – lets you run natural-language CRUD commands against any NocoDB base.
https://github.com/edwinbernadus/nocodb-mcp-serverStop switching between your AI assistant and database admin panels. This MCP server lets you run natural language commands directly against your NocoDB bases—no more clicking through interfaces or writing SQL queries.
You're using NocoDB because it turns your database into a collaborative workspace. But you're still context-switching between your AI assistant for analysis and your database interface for data operations. Every time you need to update records, check data, or bulk modify entries, you're breaking your flow.
With the NocoDB MCP server, your AI assistant becomes your database interface. Ask Claude to pull user data, update records, or even create tables from JSON files—all in natural language, all without leaving your conversation.
"get data from nocodb, table: Users where status is active"
"add new row with name: john-doe, email: [email protected]"
"update all rows, set status to verified where created_date > 2024-01-01"
"delete all rows with name containing 'test'"
That's it. No API calls, no admin panel, no SQL.
Bulk Operations Made Simple: Instead of selecting hundreds of rows manually, just describe what you want. "Update all users from the marketing campaign to VIP status" becomes a single command.
Data Analysis in Context: Pull database insights directly into your AI conversations. Ask for user counts, revenue summaries, or trend analysis without switching tools.
JSON-to-Database Pipeline: Drop a JSON file into your conversation and tell it to create a NocoDB table. Perfect for quickly setting up new datasets or migrating data.
Column Management: Add, modify, or remove database columns through conversation. "Add an 'age' column and set everyone to 25" works exactly as you'd expect.
Rapid Prototyping: Building a new app? Describe your data structure in natural language and watch your database schema materialize instantly.
Data Cleanup: Those messy imported datasets become manageable. "Remove all duplicate emails and standardize phone number formats" becomes a conversation, not a scripting project.
Client Demos: Show clients how their data looks in real-time. Pull live metrics, update sample data, and answer questions without leaving the presentation flow.
Team Collaboration: Non-technical team members can request data changes through shared AI conversations, eliminating the communication overhead of database requests.
Install and build:
git clone https://github.com/edwinbernadus/nocodb-mcp-server.git
cd nocodb-mcp-server
npm install && npm run build
Configure your NocoDB connection:
NOCODB_URL=https://your-nocodb-instance.com
NOCODB_API_TOKEN=your_api_token
NOCODB_BASE_ID=your_base_id
Connect to Claude Desktop:
{
"mcpServers": {
"nocodb": {
"command": "node",
"args": ["path/to/dist/start.js"],
"env": {
"NOCODB_URL": "https://your-nocodb-instance.com",
"NOCODB_BASE_ID": "your_base_id",
"NOCODB_API_TOKEN": "your_api_token"
}
}
}
}
This isn't just a proof of concept—it's production-ready TypeScript with proper error handling and comprehensive CRUD operations. The server handles everything from simple record lookups to complex bulk operations with built-in validation.
You can run it standalone for testing (npx nocodb-mcp-server), integrate it with Claude Desktop for daily use, or embed it in your own MCP-compatible tools.
Your NocoDB database just became conversational. Time to put it to work.