filemaker-mcp

Connect Claude (or any MCP client) to a FileMaker database — read-only queries, schema discovery, and pandas-powered analytics.

What It Does

filemaker-mcp is an MCP server that gives AI assistants live access to your FileMaker data via OData v4. Load it in Claude Desktop or Claude Code and ask questions about your data in plain English.

Tools provided:

• fm_query_records — Search and filter records with OData expressions

• fm_get_record — Fetch a single record by primary key

• fm_count_records — Count records with optional filters

• fm_list_tables — List available tables

• fm_get_schema — Discover field names, types, and keys

• fm_load_dataset — Pull records into memory for analytics

• fm_analyze — Run groupby/sum/count/mean/min/max on loaded data

• fm_list_datasets — See what datasets are loaded

Quick Start

Prerequisites

• Python 3.12+

• uv package manager

• FileMaker Server with OData v4 enabled

• An FM account with fmodata extended privilege

Install

git clone https://github.com/nietsneuah/filemaker-mcp.git
cd filemaker-mcp
cp .env.example .env
# Edit .env with your FileMaker server details
uv sync

Configure Claude Desktop

Add to your Claude Desktop MCP config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "filemaker": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/filemaker-mcp", "filemaker-mcp"],
      "env": {
        "FM_HOST": "your-server.example.com",
        "FM_DATABASE": "your_database",
        "FM_USERNAME": "mcp_agent",
        "FM_PASSWORD": "your_password"
      }
    }
  }
}

Run

uv run filemaker-mcp

Schema Discovery

On startup, the server auto-discovers your tables from the OData service document. For richer schema (field types, primary keys, tiers), install the optional GetTableDDL FileMaker script — see docs/FM_ACCOUNT_SETUP.md.

Analytics

For reports and summaries, use the analytics tools instead of raw queries:

1. fm_load_dataset — Fetch records into a pandas DataFrame (auto-paginates)

2. fm_analyze — Run aggregations instantly (no additional FM round trips)

This returns ~200 tokens instead of ~400K for raw records — much more efficient for dashboards and trend analysis.

Documentation

• Architecture — System design and security model

• FM Account Setup — FileMaker privilege configuration

• Tool Reference — Detailed parameter docs

License

GPL-3.0 — see LICENSE

Author

Doug Hauenstein / FM Rug Software