Connect an AI agent to FileMaker without handing it the keys.
Your FileMaker system holds the business knowledge an AI agent needs. FileMaker Server can share that data over JDBC, so an agent can ask questions, build reports, and explore your schema using standard SQL.
BasicData sets up the connection properly: a dedicated read-only account, production protected by default, writes gated behind approval, and every query logged. You get the benefit of an AI agent without the risk.
What connecting an AI agent to FileMaker actually involves.
Turning on JDBC is the easy part. A connection an agent can use safely needs scoped accounts, query guardrails, FileMaker-specific SQL knowledge, and an audit trail. We build all of it so the agent is useful on day one and safe long after. Where the official Claris MCP server is the better fit, we set that up instead and apply the same guardrails.
JDBC and ODBC sharing enabled and locked down on FileMaker Server
Dedicated read-only database account for the agent (fmxdbc extended privilege)
Agent connection layer: a custom MCP server or tool functions over JDBC, or Claris MCP where it fits
SQL query tooling tuned for FileMaker SQL quirks and limits
Schema discovery so the agent learns your tables and fields safely
Gated write access with explicit approval for inserts, updates, deletes
Row limits, query timeouts, and protection against slow unstored-field scans
Audit logging of every query the agent runs, at the connection layer
Development and staging connections so production is never the test bed
Documentation and handover for your team
An AI agent with the wrong access is a risk, not an asset.
Most of the risk in agent projects sits in the data connection, not the model. An agent running on an administrator login, with write access to production and no logging, will eventually do something nobody can explain. The fix is design: least-privilege accounts, read-only by default, and a human in the loop for anything that changes data.
FileMaker SQL is not standard SQL, and an agent that does not know the rules writes queries that fail or hammer your server. JDBC can also write and even change schema, and those changes bypass script triggers and layout-level logic. Read-only must be enforced by the account's privilege set, not by a prompt. We build tooling that teaches the agent how FileMaker behaves: identifier quoting, empty strings treated as NULL, case-sensitive text matching, the unsupported functions, and the slow query patterns to avoid. We run this setup on our own FileMaker systems every day.
We are a registered member of the Claris FileMaker Business Alliance, the formal partner programme for professional FileMaker development companies.
Businesses that want answers from their FileMaker data without building a new system.
Operations teams who want to ask their own data questions in plain English. Managers who want reporting that does not wait on a developer. Teams who have inherited an undocumented FileMaker solution and need to understand it fast. Developers who want an AI coding assistant that can read their schema. If your data lives in FileMaker and you want an agent working with it, this is the integration.
Questions about AI agents, JDBC and MCP.
How agents connect, what they can do, and how to keep your data safe.
Can an AI agent connect directly to FileMaker?
Yes. FileMaker Server can share hosted files over JDBC and ODBC, so any agent that can run SQL can query FileMaker. That includes an MCP server, a tool-calling function, or a Python or Java service. The agent talks to the server, never to the FileMaker client.
What is the FileMaker JDBC driver?
It is the JDBC client driver Claris provides for connecting Java-based applications to FileMaker using SQL. You enable ODBC/JDBC in the FileMaker Server Admin Console under Connectors, and give the account a privilege set with the Access via ODBC/JDBC (fmxdbc) extended privilege.
What is the difference between JDBC and Claris MCP?
Claris MCP is the official MCP server from Claris. It works through the Data API and OData, needs a recent FileMaker Server 22 release and a Claris ID, and exposes the tables and scripts you choose. JDBC gives an agent flexible SQL across tables, which suits questions, reporting and schema exploration. We recommend the right one for your setup, and sometimes both.
Is it safe to give an AI agent access to my FileMaker data?
It can be, if access is designed properly. The agent gets its own read-only account scoped to the tables it needs. Production writes are off by default, result sizes are capped, and every query is logged. Any change to data is a separate, human-approved step. The agent never receives administrator credentials.
Can an AI agent write to FileMaker over JDBC?
Technically yes. JDBC supports INSERT, UPDATE and DELETE, and can even alter tables. That is why read-only has to be enforced by the account privileges, not by instructions to the agent. Writes over SQL also skip script triggers, so where a change must run business logic we route it through the Data API or a FileMaker script instead.
Should the agent use JDBC or the FileMaker Data API?
They suit different jobs. JDBC is best for reading: answering questions, reporting and exploring data. The Data API and OData work through layouts and can run FileMaker scripts, which makes them better when a change must trigger validation or scripted behaviour. A common pattern is JDBC to read and the Data API to act.
What are the limits of FileMaker SQL for AI agents?
FileMaker supports a subset of SQL, documented in the Claris FileMaker SQL Reference. Several common functions are missing, identifiers often need quoting, and SQL cannot run scripts or create calculation fields. Filtering unstored or unindexed fields on large tables is slow. Good tooling teaches the agent these rules up front.
Do I need FileMaker Server, and which version?
For practical use, yes. FileMaker Pro can only share a file over ODBC/JDBC to the same computer, with a handful of connections. FileMaker Server supports remote connections. Claris MCP additionally needs a recent FileMaker Server 22 release. We review your version and hosting first and tell you if anything needs upgrading.
Can I connect Claude or ChatGPT to FileMaker?
Yes. The approach is model-agnostic. Any assistant or agent platform that supports MCP or SQL tools, including Claude and ChatGPT-based agents, can use the same secured connection. We build the data layer once, and you choose the model.
Tell us what you want your AI agent to do with FileMaker.
Describe the questions you want answered or the tasks you want automated. We will tell you what the right connection looks like and what it involves.