Connect Claude to your charity's tools with MCP

The problem

You're copying and pasting data between Claude and your other tools constantly. Pull a report from your CRM, paste it into Claude, ask a question, copy the answer back. Or you want Claude to look something up in your database but have to manually fetch and paste the data first. Every conversation requires this tedious shuttle of information.

The solution

Use Model Context Protocol (MCP) to connect Claude directly to your charity's data sources and tools. MCP is an open standard that lets AI models talk to external systems - databases, APIs, file systems, even platforms like Slack or Google Drive. Once connected, Claude can query your CRM, read your shared documents, or check your project management tool without you doing the copy-paste dance.

What you get

Claude Desktop or Claude Code connected to one or more of your charity's systems via MCP servers. You can ask Claude questions that draw on live data: 'How many referrals did we get last month?', 'Summarise the latest board minutes from our shared drive', 'What grants are due for reporting this quarter?' The data stays in your systems; Claude reads it on demand.

Before you start

• Claude Desktop app or Claude Code installed
• Claude Pro subscription or above
• Technical confidence: you need to edit configuration files and potentially run commands in a terminal
• Node.js or Python installed on your computer
• Access credentials for the systems you want to connect (API keys, database passwords)
• CRITICAL: Understand that MCP gives Claude read access to your systems. Only connect data sources where this is appropriate. Do not connect systems containing unencrypted beneficiary data without a data protection assessment.

When to use this

• You're constantly copying data from other systems into Claude
• You want Claude to answer questions using live data from your CRM or database
• You have a technical team member who can set up and maintain the connections
• You want to build more sophisticated AI workflows that draw on multiple data sources

When not to use this

• You have no one with technical skills to set up and maintain MCP servers
• Your data is too sensitive to expose to an external AI (safeguarding records, identified beneficiary data)
• Your needs are met by uploading files to Claude manually
• Your systems don't have APIs or database access available

Steps

1. 1

   Identify what to connect

   List the systems where you repeatedly copy data from to paste into Claude. Common examples: Google Drive (for documents), your CRM or database (for records), Slack (for messages), project management tools. Start with one system - don't try to connect everything at once.

2. 2

   Find or build an MCP server

   Check the MCP server registry at modelcontextprotocol.io for pre-built servers. There are servers for Google Drive, Slack, PostgreSQL, SQLite, file systems, and many more. For Blackbaud, Candid, and Benevity, Anthropic has released official nonprofit MCP connectors. If your system isn't covered, you'll need to build a custom server using the MCP SDK.

3. 3

   Configure Claude Desktop

   Add the MCP server to your Claude Desktop configuration file (claude_desktop_config.json on Mac/Windows). Each server needs its connection details: the command to start it, environment variables for API keys, and any configuration options. Claude's documentation walks through this step by step.

4. 4

   Test with safe data first

   Restart Claude Desktop and verify the MCP connection appears (you'll see available tools in the interface). Test with non-sensitive queries first: 'List the files in our shared drive', 'Show me last month\'s donation totals'. Check that the data coming through is accurate and complete.

5. 5

   Set access boundaries

   MCP servers can be configured to limit what Claude can access. Set read-only access where possible. Restrict which database tables or folders are exposed. Document what Claude can and can't access. This is your data protection boundary.

6. 6

   Document and share with the team

   Write up what's connected, what Claude can now do, and any limitations. If other staff will use this, they need to understand what data Claude can access and the same data protection considerations that apply to any tool with access to your systems.

Tools

Resources