mcp-server-filemaker
This server lets MCP-compatible AI assistants read and manage FileMaker databases on-premise over OData v4, including CRUD, schema introspection, scripting, batch queries, and test/validation tools.
Query & read data: list tables, fetch full metadata/EDMX, introspect tables/fields, query with $filter/$select/$top/$orderby/$expand/$count, and fetch single records by ROWID.
Create, update, delete records: insert new records, patch existing ones, and delete records.
Run FileMaker scripts: execute scripts via OData, optionally with parameters, and assert the result code.
Modify schema: create new tables and add fields with SQL-style types (VARCHAR, INT, NUMERIC, DATE, etc.).
Batch operations: combine multiple OData GET requests into a single HTTP call.
Data API bridge: set global fields and run scripts in the same session context (requires fmrest privilege).
Testing & validation: create tagged test records, clean up test data, assert record field values, and assert query result counts.
Operational control: restrict dangerous tools per database via FM_DISABLED_TOOLS, configure timeout/retry, and enable debug logging when needed.
Provides tools for managing FileMaker databases via OData v4 API, including CRUD operations, script execution, schema introspection, batch requests, and test/validation tools.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-server-filemaker List customers from New York"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mcp-server-filemaker
MCP server that gives AI assistants direct access to FileMaker databases via OData v4 API.
Works with any MCP-compatible client: Claude Code, Claude Desktop, Cursor, Windsurf, Continue.dev, Zed, and others.
Runs on-premise — no cloud proxy, no Claris ID required.
MCP Client ──stdio──▶ MCP Server (Node.js) ──OData v4 / HTTPS──▶ FileMaker ServerFeatures
20 tools — CRUD, scripts, schema introspection, batch requests, test & validation
OData v4 — direct table access (not layout-based like the Data API)
Data API bridge — set global fields and run scripts with session context
Tool restrictions — disable dangerous tools per database via
FM_DISABLED_TOOLSZero dependencies beyond the MCP SDK and Zod
Related MCP server: filemaker-odata-mcp
Prerequisites
FileMaker Server with OData access enabled (Admin Console → Connectors → OData)
A dedicated API user with Extended Privilege
fmodataFor Data API tools (
fm_set_globals,fm_run_script_with_globals): alsofmrestHTTPS with a valid certificate — self-signed certificates are not supported
Installation
git clone https://github.com/JoergKoester/mcp-server-filemaker.git
cd mcp-server-filemaker
npm install
npm run buildConfiguration
Add the server to your MCP client config. For Claude Code and Claude Desktop, edit ~/.mcp.json:
{
"mcpServers": {
"filemaker": {
"command": "node",
"args": ["/absolute/path/to/mcp-server-filemaker/dist/index.js"],
"env": {
"FM_HOST": "https://your-filemaker-server.example.com",
"FM_DATABASE": "YourDatabase",
"FM_USERNAME": "api_user",
"FM_PASSWORD": "your_password",
"FM_ODATA_VERSION": "v4",
"FM_TIMEOUT_MS": "15000",
"FM_RETRY_COUNT": "2",
"FM_DISABLED_TOOLS": "",
"FM_DEBUG": "false"
}
}
}
}Environment Variables
Variable | Required | Description |
| Yes | FileMaker Server URL with protocol, e.g. |
| Yes | Database name |
| Yes | API user (needs Extended Privilege |
| Yes | Password |
| No | OData version, default |
| No | HTTP request timeout in milliseconds, default |
| No | Retry attempts on transient failures (429/503), default |
| No | Size limit for |
| No | Comma-separated tool names to disable |
| No |
|
Multiple Databases
Run one server instance per database:
{
"mcpServers": {
"filemaker-app": {
"command": "node",
"args": ["/path/to/dist/index.js"],
"env": {
"FM_HOST": "https://fms.example.com",
"FM_DATABASE": "MyApp",
"FM_USERNAME": "api_user",
"FM_PASSWORD": "secret",
"FM_DISABLED_TOOLS": "fm_create_table,fm_add_field"
}
},
"filemaker-data": {
"command": "node",
"args": ["/path/to/dist/index.js"],
"env": {
"FM_HOST": "https://fms.example.com",
"FM_DATABASE": "MyData",
"FM_USERNAME": "api_user",
"FM_PASSWORD": "secret",
"FM_DISABLED_TOOLS": "fm_create_table,fm_add_field,fm_delete_record"
}
}
}
}Tools
Structure & Metadata
Tool | Description |
| List all tables |
| Full EDMX schema (fields, types, relationships) |
| Tables overview or field details with native FM types |
Data Access
Tool | Description |
| Query records with |
| Single record by key (see below) |
| Create a new record |
| Update record (PATCH) |
| Delete record by key (see below) |
Addressing a single record. FileMaker declares an OData key only for tables whose key field is validated as unique and not empty; those tables are addressed by that field's value. Every other table is addressed by its numeric
ROWID, which is returned with each query result. Pass either form asrowId— quoting is handled by the server.
Scripts & Schema
Tool | Description |
| Execute a FileMaker script |
| Create a new table |
| Add a field (SQL-style types: |
| Batch multiple OData requests in one HTTP call |
Data API (Session Context)
Tool | Description |
| Set global fields (Login → PATCH /globals → Logout) |
| Set globals + run script in one session |
Test & Validation
Tool | Description |
| Create a tagged test record |
| Delete all records matching a test tag |
| Validate field values (PASS/FAIL with diff) |
| Assert record count for a query |
| Run script and check result code |
Tool Restrictions
Disable tools per database instance via FM_DISABLED_TOOLS:
FM_DISABLED_TOOLS=fm_create_table,fm_add_field,fm_delete_recordDisabled tools are hidden from the tool list and blocked at execution (double protection)
Server startup shows
Tools: 17/20and lists disabled toolsRecommended: always disable
fm_create_tableandfm_add_fieldon production databases
Architecture
4 files in src/:
File | Responsibility |
| Entry point, env config, STDIO transport, tool routing |
| HTTP client for all OData v4 requests (CRUD, scripts, schema, batch) |
| HTTP client for FileMaker Data API (globals, session-scoped scripts) |
| MCP tool definitions (Zod schemas), handler logic, test tools |
The server is a thin proxy — no caching. Each OData request uses Basic Auth; Data API tools manage their own session (login → action → logout). The OData client retries transient failures (HTTP 429/503) with exponential backoff; the Data API client currently does not retry.
Security Notes
stdio transport has no authentication of its own — anyone who can start the process has full access to the configured FileMaker database with the rights of the API user. Credentials live in your MCP client config (e.g.
~/.mcp.json) and are protected by filesystem permissions only. In multi-user environments, secure the server process and config file accordingly.Startup logs to stderr include the configured host, database, and username (no password, no API keys). Redirect stderr appropriately if that is sensitive.
FM_DEBUG=truelogs the full arguments of every tool call to stderr, which may contain record data from your FileMaker database. Keep this off in production.OData
$filtervalues passed tofm_queryare placed verbatim into the query string. Escape single quotes ('→'') in string literals yourself, and be aware that values containing URL-reserved characters (#,+,%) can corrupt the request.
Development
npm run dev # Run with tsx (no build needed)
npm run watch # Watch mode
npm run build # Compile TypeScript → dist/
npm start # Run compiled versionOData API Reference
Operation | Method | Endpoint |
Service Document | GET |
|
Metadata (EDMX) | GET |
|
Read records | GET |
|
Create record | POST |
|
Update record | PATCH |
|
Delete record | DELETE |
|
Run script | POST |
|
Create table | POST |
|
Add field | PATCH |
|
Batch | POST |
|
AI-generated experimental project
This MCP server was generated almost entirely with the assistance of a generative AI system and has only undergone limited human review so far. It is provided as is, without any guarantees of correctness, security, or fitness for a particular purpose. Use it at your own risk and always review and test the code thoroughly before using it in production.
The code in this repository is licensed under the MIT License. By using it, you agree that the authors and copyright holders are not liable for any claim, damages, or other liability arising from its use, as stated in the license.
License
MIT
Available Tools
19 toolsfm_add_fieldC
Add a new field to an existing FileMaker table. SQL-style field types: VARCHAR(n), INT, NUMERIC, DATE, TIME, TIMESTAMP, BLOB.
Inputs: table (required), fieldName (required), fieldType (required), fieldRepetition.
fm_assert_countB
Checks whether a query returns the expected number of records. Useful for validating script results.
Inputs: table (required), expected (required), filter.
fm_assert_recordA
Validates that a FileMaker record contains the expected field values. Returns PASS/FAIL with diff.
Inputs: table (required), rowId (required), expected (required).
fm_batchA
Execute multiple OData GET requests in a single HTTP call (batch). Use individual tools for write operations.
Inputs: requests (required).
fm_cleanup_test_dataB
Deletes all test records marked with a specific tag. tagField must be an existing text field in the FM table.
Inputs: table (required), tagField (required), tag.
fm_create_recordC
Create a new record in a FileMaker table.
Inputs: table (required), fields (required).
fm_create_tableB
Create a new table in the FileMaker database (schema modification via OData).
Inputs: tableName (required).
fm_create_test_recordA
Creates a test record with a test tag. tagField must be an existing text field in the FM table. Returns ROWID for later validation and cleanup.
Inputs: table (required), fields (required), tagField (required), tag.
fm_delete_recordC
Delete a FileMaker record by ROWID.
Inputs: table (required), rowId (required).
fm_get_metadataB
Reads the full OData $metadata document of the FileMaker database (EDMX/XML). Contains all tables, fields, types, and relationships.
Inputs: no named parameters.
fm_get_recordB
Read a single FileMaker record by ROWID.
Inputs: table (required), rowId (required).
fm_get_service_documentA
Lists all available EntitySets (tables/layouts) of the FileMaker database.
Inputs: no named parameters.
fm_introspectA
Analyzes the database structure: lists tables or shows all fields of a table with native FM field types from $metadata.
Inputs: table.
fm_queryA
Query records from a FileMaker table. Supports OData $filter, $select, $top, $skip, $orderby, $expand, $count. Default limit: 100 records (override with $top).
Inputs: table (required), filter, select, top, +4 more.
fm_run_scriptC
Run a FileMaker script. Returns scriptResult with code and resultParameter.
Inputs: scriptName (required), scriptParam.
fm_run_script_and_assertB
Runs a FileMaker script and checks whether the result code matches the expected value.
Inputs: scriptName (required), scriptParam, expectedResultCode.
fm_run_script_with_globalsA
Sets global fields and runs a script in the SAME Data API session. This makes scripts that depend on global fields for context work (e.g. SESSIONS::UUID_g). Flow: Login → Set Globals → Run Script → Logout.
Inputs: globals (required), layout (required), scriptName (required), scriptParam.
fm_set_globalsA
Sets global fields via FileMaker Data API (not OData). Opens a Data API session, sets the globals, then closes the session. Keys must be fully qualified: 'Table::FieldName_g'.
Inputs: globals (required).
fm_update_recordC
Update an existing FileMaker record by ROWID (PATCH).
Inputs: table (required), rowId (required), fields (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 19 tools
Each tool serves a distinct purpose: CRUD, metadata, query, testing, scripting, schema, batch. Overlaps are intentional (e.g., create_record vs create_test_record). No ambiguity between tools.
All tools follow a consistent 'fm_verb_noun' pattern. Verbs are clear and standardized (create, get, update, delete, assert, run, set, add, cleanup, introspect). Minor deviations like 'batch' are still effective.
19 tools is slightly above the typical 3-15 range, but each tool addresses a specific aspect of FileMaker interaction (CRUD, schema, testing, scripting, metadata). The count feels earned and not excessive.
Core operations are well-covered: CRUD, query, metadata, scripting, testing. Minor gaps include lack of schema modification beyond adding fields (e.g., no delete table/field) and no tool for file uploads, but these are non-essential.
Maintenance
Related MCP Connectors
Cloud-hosted MCP server for secure AI access to enterprise data sources via CData Connect AI.
MCP server unifying ERPs, CRMs, APIs and knowledge base for Claude, ChatGPT and Gemini.
MCP server that lets AI assistants use all OneSchema features exposed via the public API.
Hosted MCP server connecting AI assistants to 9,000+ apps and 40,000+ actions via Zapier.
Related MCP Servers
- AlicenseAqualityBmaintenanceMCP server for Creatio CRM that enables AI assistants to read, create, update, delete records, execute business processes, and manage settings via OData and REST APIs.2477 npm6MIT
- AlicenseNot gradedqualityBmaintenanceMCP server for FileMaker Server OData 4.01 API integration, enabling AI assistants to discover databases, perform CRUD operations, and manage connections.21 npm14MIT
- AlicenseNot gradedqualityDmaintenanceAn MCP server that enables AI assistants to interact with Odoo ERP, allowing natural language queries, record creation, updates, and deletions.LGPL 3.0
- AlicenseNot gradedqualityCmaintenanceAn MCP server that enables AI assistants to interact with Odoo ERP systems, allowing natural language access to business data, CRUD operations, and instance management without requiring Odoo module installation.1MIT