ZeroBlue
Blog •

FileMaker and AI: The Future Is Already Here

For decades, I have watched Claris FileMaker quietly power the operations of hospitals, universities, government agencies, nonprofits, and businesses of every size — not because it was the loudest tool in the room, but because it worked. It adapted. It stayed in the hands of the people who actually understood the problem being solved. Now, as artificial intelligence reshapes the software landscape, I think FileMaker is in an unexpectedly strong position: a mature, flexible, data-rich platform sitting at the exact intersection where AI does its best work.

This is my view of what’s happening right now, what it means for FileMaker developers and the organizations they serve, and where I think the ecosystem is heading.

A Platform With Deep Roots

Before we talk about the future, I think it’s worth appreciating the foundation. FileMaker — now officially branded as Claris FileMaker — has been in continuous development since 1985. It has shipped over 16 million copies to organizations spanning large enterprises, small businesses, school districts, and government agencies. That’s not a niche tool. That’s infrastructure.

The platform’s sweet spot has always been custom application development without requiring a traditional software engineering team. A smart operations manager, a research coordinator, a clinic administrator — people who understand their workflow deeply but aren’t career programmers — could build sophisticated, relational database applications that actually fit how their organization worked. That philosophy, of putting powerful technology into the hands of problem solvers rather than gatekeeping it behind specialized engineering teams, turns out to align almost perfectly with where AI is going.

Who’s Actually Using It

The breadth of FileMaker’s real-world deployment is easy to underestimate, even for people who have worked around it for years. From Claris’s own published case studies, the platform powers field research data management for the Great Maya Aquifer Project, post-seismic risk assessment for real estate firms, equipment tracking for production lighting companies, and conservation monitoring for elephant researchers in India. Travis County Fire Rescue in Texas built a vaccine tracker on FileMaker to accelerate COVID-19 distribution. Beverly Public Schools used it to manage remote learning logistics during the pandemic. OceanX — the ocean exploration organization — keeps scientists, crew, and data synchronized across land and sea with it.

In academia, large universities have deployed FileMaker across multiple departments for everything from research data management to HR and scheduling, often consolidating dozens of fragmented departmental databases under a single platform. The U.S. government and associated agencies have long been FileMaker users — the platform’s security model, role-based access controls, and Active Directory integration make it a reasonable fit for sensitive operational environments where custom commercial software is unavailable or inappropriate.

What these deployments share is a common characteristic: they are rich with domain-specific, structured, often proprietary data that lives nowhere else. And that is precisely what makes them such fertile ground for AI.

FileMaker 2025: A Strategic Pivot to AI

When Claris released FileMaker 2025 in July of that year, it did not feel incremental to me. Claris CEO Ryan McCann described the moment plainly: “We’re beyond AI being a trend — it’s reshaping how we work and the tools we choose today.” The release marked a deliberate repositioning of the platform as a first-class environment for building AI-powered business applications.

The headline AI features in FileMaker 2025 represent a coherent stack rather than a collection of loosely related additions:

Generate Response from Model — A new script step that lets developers send prompts directly to an AI model and receive text responses, right inside FileMaker’s scripting environment. This enables automated report summarization, dynamic customer communications, intelligent document templates, and on-demand content generation without any external tooling.

Natural Language Queries — Users can now ask questions about their FileMaker data in plain English and receive answers, converting natural language prompts into SQL or FileMaker-native queries behind the scenes. For organizations where non-technical staff need to surface insights from complex relational databases, this is transformative.

RAG (Retrieval-Augmented Generation) — FileMaker 2025 introduced built-in support for RAG systems, allowing developers to build context-aware AI assistants that generate answers grounded in the organization’s own documents — PDFs, manuals, policy documents, research notes. The AI doesn’t just guess; it reasons from your data. A school district, for example, could deploy an internally hosted AI assistant that answers HR-related employee questions by referencing its own policy documents — exactly the kind of use case Claris has highlighted.

Vector Embeddings and Semantic Search — Text and images can now be converted into embedding vectors stored in FileMaker container fields, enabling semantic search across both textual and visual data. This opens the door to finding records by meaning rather than keyword matching — a meaningful upgrade for research databases, media libraries, and knowledge management systems.

Bring Your Own Model — Critically, FileMaker 2025 doesn’t lock you into a single AI provider. Developers can configure integrations with OpenAI, Anthropic’s Claude, Cohere, or open-source local LLMs. For organizations with data privacy requirements — healthcare providers, government contractors, academic institutions handling sensitive research data — the ability to run models on local infrastructure without data leaving the organization’s control is not a nice-to-have; it’s a compliance requirement.

AI Model Server in FileMaker Server — The server-side companion to these features is equally important. FileMaker Server 2025 now includes a built-in AI Model Server manageable directly from the Admin Console, supporting open-source models for text generation, text embedding, and image embedding. Administrators can download, load, fine-tune, and manage models from the same interface they already use to manage their databases. Organizations can even fine-tune models on their own JSONL training data — effectively creating AI that understands their specific domain vocabulary, workflows, and data patterns.

The MCP Bridge: Connecting FileMaker to the Broader AI Ecosystem

If the 2025 AI features represented FileMaker gaining intelligence from the inside out, Claris MCP (Model Context Protocol) — introduced in December 2025 — represents something complementary and arguably more profound: FileMaker becoming a data source and action surface for external AI agents.

MCP is an open standard originally developed by Anthropic to create a standardized way for AI models to interact with external data sources and tools. Rather than requiring a custom integration for every AI client, MCP provides a universal protocol — a shared language between AI systems and the applications they interact with.

Claris’s implementation of this standard turns FileMaker Server into an MCP Server. Once configured, AI assistants like Claude can connect to a hosted FileMaker database, understand its schema, query records, update data, and trigger FileMaker scripts — all through natural language conversation, with no additional API development required. The setup requires FileMaker Server 22.0.2.204 or newer, with OData and the Data API enabled, and takes minutes to configure through the Claris AI Workspace console.

This is also where FM Dojo AI Data Access fits in. Instead of exporting FileMaker data to CSV, pasting records into chat, or putting FileMaker credentials directly into local AI configuration files, FM Dojo gives an AI tool a scoped, revocable access path to approved FileMaker databases. You connect the database in FM Dojo, generate a token or connector URL, and then use that connection from Claude Desktop, Claude.ai, Claude Mobile, Cursor, Zed, Continue.dev, or another MCP-compatible client.

For Claude specifically, the workflow is practical. Claude Desktop can use the MCP connection so you can ask questions against live FileMaker data from your Mac. Claude.ai and Claude Mobile can use the same idea through a connector URL, which matters because Claude supports a limited custom connector workflow; one FM Dojo token can cover multiple FileMaker databases when you want Claude to reach them through a single connection. The AI tool receives the FM Dojo token, not your FileMaker username and password, and access can be revoked without changing the database credentials.

I recorded a quick walkthrough showing this setup with Claude, including how it works from desktop, web, and mobile.

Watch the walkthrough on YouTube.

The implications are significant. An organization running decades of operational data in FileMaker can now give an AI agent like Claude access to that data in a controlled, auditable, permission-aware way. As a developer, I can ask: “Which customers haven’t placed an order this year?” and receive an answer pulled from live FileMaker data. A manager can say: “Show me all overdue invoices, then for anything older than 30 days, mark them as high priority and notify the billing team” — and watch it happen, step by step, using existing FileMaker scripts and business logic that the organization has already validated.

Crucially, MCP doesn’t bypass FileMaker’s security model. AI access is granted through standard FileMaker accounts with their existing privilege sets. The AI can only interact with tables and fields that account can access. All actions are logged through FileMaker Server’s standard audit trail. This matters enormously for regulated industries and compliance-conscious organizations.

The Developer Ecosystem: Tools That Meet You Where You Are

Claris doesn’t build in a vacuum, and the developer ecosystem surrounding FileMaker has always been one of its quiet strengths. I have seen that ecosystem from several angles over the years, including as a developer, consultant, and former Claris employee. Firms like Soliant Consulting, DB Services, Kyo Logic, The Support Group, Skeleton Key, Codence, and many others have built deep practices around the platform, contributing open-source tooling, community resources, and production-grade expertise that makes the ecosystem more capable than any single vendor could achieve alone.

In the AI era, this ecosystem is moving quickly.

FM Dojo is purpose-built AI tooling that grew out of requests I kept hearing from the FileMaker community. I built it from more than 30 years of FileMaker experience, including time at Claris (Apple), Lawrence Livermore National Laboratory, and Soliant Consulting. FM Dojo is an AI-powered development assistant designed specifically for the FileMaker platform.

The problem FM Dojo solves is one I kept running into myself: general-purpose AI tools like ChatGPT or stock Claude don’t understand FileMaker’s scripting language, calculation engine, ExecuteSQL syntax, or the specific quirks of how the platform behaves. They hallucinate function names, suggest script steps that don’t exist, and produce advice that sounds reasonable but breaks in practice. FM Dojo addresses this by training its AI specifically on FileMaker’s actual behavior — testing edge cases against real FileMaker instances, building strict argument catalogs for calculation functions, and enforcing the platform’s known constraints.

The platform has grown rapidly since its February 2026 launch, shipping over 30 releases in its first months. I have tried to keep the feature set grounded in what working FileMaker developers actually need:

  • AI Chat — context-aware assistance for scripts, calculations, ExecuteSQL, JSON functions, and database design, with full understanding of FileMaker’s syntax and edge cases
  • SQL Fiddle — run ExecuteSQL-compatible queries against a mock schema in real time to validate logic before deploying
  • Script Builder — write, validate, and export FileMaker scripts with IDE-quality autocomplete for script steps, functions, and your own schema fields
  • Schema Analyzer — upload your database’s XML export for AI-powered analysis covering table structure, naming conventions, anchor-buoy relationship detection, and best-practice audits
  • MCP Server — FM Dojo includes its own MCP server, meaning AI tools like Claude Desktop, Cursor, and others can connect to live FileMaker data and answer questions or take actions in natural language
  • Flows — a Claris Connect alternative for connecting FileMaker to external web services, with over 90 integrations, enabling automation without writing middleware
  • Server Deploy — tooling for managing FileMaker Server deployments
  • Server Monitoring — ongoing observability for hosted FileMaker environments

FM Dojo’s token-based pricing model (starting at $20/month for 350K tokens, with unused tokens rolling over) is my attempt to align cost with actual usage rather than forcing flat subscriptions on developers whose workloads vary. A free trial tier with a 200K token cap lowers the barrier to evaluation.

To me, this is what happens when deep platform expertise meets modern AI tooling: not a generic wrapper around a language model, but a system that actually understands the domain it serves.

Real-World AI Applications in FileMaker Today

The abstract potential of AI integration becomes more concrete when you look at the kinds of applications the platform is being positioned to support. Claris has published a catalog of FileMaker AI use cases that illustrates where the platform fits, ranging from the practical to the aspirational:

  • An internally hosted AI assistant for a school district that answers HR-related questions by referencing the organization’s own policy documents through RAG
  • Call center workflows where AI generates suggested responses for customer service reps, drawing context from live customer records in FileMaker
  • HR tools that match applicant resumes against open positions using semantic similarity rather than keyword matching
  • On-demand sales reporting and analytics where AI summarizes FileMaker data into narrative form
  • Customer review categorization that routes feedback and surfaces themes from incoming text
  • AI-assisted quality control and documentation workflows in engineering and manufacturing contexts

I think it’s worth being honest about what this list represents. Some of these are documented production deployments; others are reference architectures or early implementations that Claris is putting forward as templates for what the platform now supports. The 2025 AI features and Claris MCP are genuinely new — most organizations adopting them are doing so in the first year of availability, which means we’re still in the early-deployment phase rather than the mature-best-practices phase. I expect the most compelling case studies to emerge over the next 12 to 24 months as these implementations mature in production.

In healthcare, FileMaker has long been used for clinical data management, research tracking, and patient workflow coordination. AI adds the ability to summarize clinical notes, surface anomalies in research data, and automate routine documentation — all while keeping sensitive data on controlled infrastructure.

In academic research — where FileMaker has a presence in everything from archaeological fieldwork to marine biology to cancer research — AI opens the door to literature-aware data entry, automated tagging and categorization, and natural language querying of complex relational research databases.

In government and public sector contexts, where custom commercial software often doesn’t exist for specialized workflows, FileMaker’s AI capabilities offer a way to bring modern intelligence to legacy operational data without a full platform migration.

Honest Limitations and Open Questions

A realistic assessment of FileMaker’s AI trajectory has to acknowledge the complexity involved. I am excited about where this is going, but I also see a few genuine open questions.

The AI Model Server has real infrastructure requirements. Running local LLMs requires GPU-accelerated hardware (Apple Silicon or CUDA-enabled NVIDIA GPUs). Organizations running FileMaker on commodity Windows servers or older Linux machines will need hardware investment to take full advantage of local model hosting. CPU-only model formats like GGUF are not supported.

The developer learning curve is real. Claris has made AI integration dramatically more accessible than it was two years ago, but building production-quality AI features — especially RAG pipelines and semantic search — still requires developers to understand concepts like embedding dimensions, vector similarity, context window management, and prompt engineering. The tools are available; the expertise to use them well is still developing across the community. FM Dojo can fill the gaps for developers by offering guidance and script writing.

Data quality matters more, not less. AI systems amplify the quality of the data they work with. FileMaker deployments that have accumulated years of inconsistent records, poorly structured fields, or incomplete data will find that AI surfaces those problems rather than hiding them. This is ultimately a good thing, but it’s work.

General-purpose AI tools remain limited for FileMaker. I am candid about this because it affects real developer productivity. Asking ChatGPT or a general Claude instance to help debug a complex FileMaker ExecuteSQL calculation or design a multi-table relationship schema often produces plausible-sounding but incorrect results. The platform’s specific syntax, function behaviors, and design patterns are underrepresented in general training data. Purpose-built tools like FM Dojo, and the community resources being built by ecosystem firms, exist specifically to address this gap — and they matter.

The ecosystem is still catching up. While FileMaker 2025 and Claris MCP represent genuine platform-level leaps, the broader ecosystem of training materials, certified developer expertise in AI integration, and production case studies is still maturing. Organizations moving quickly will be working somewhat ahead of documented best practices, which is both an opportunity and a risk management consideration.

What the Convergence Means

FileMaker’s core value proposition has always been that the people who understand a problem should be able to build the software that solves it, without requiring an intermediary engineering organization to translate intent into code. That is one of the reasons I have stayed with the platform for so long. AI doesn’t change that proposition — it accelerates it. When an AI assistant genuinely understands FileMaker’s scripting language, calculation engine, and relational model, it becomes something close to an expert pair programmer available to every developer regardless of their experience level.

And when AI models can connect directly to live FileMaker data through MCP — querying records, running scripts, updating fields, triggering workflows — they stop being answer machines and become operational agents. The difference between a system that tells you what to do and one that can do it for you, within the guardrails your organization has defined, is enormous.

For the universities running research data on FileMaker: imagine an AI that can answer complex cross-departmental queries in natural language, or that can automatically catalog new entries against existing records using semantic matching. For government agencies tracking field operations: imagine a system that can surface patterns, flag anomalies, and generate briefings from operational data without a data analyst in the loop for every request. For the small clinic or nonprofit tracking patient outcomes with a decade of FileMaker data: imagine an AI that can help clinicians find relevant historical precedents or surface gaps in care documentation on demand.

None of this requires rebuilding from scratch. The data is already there. The platform already has the right security model, the right deployment flexibility, and — as of 2025 — the right native AI integration layer. The question I keep coming back to is not whether organizations should pursue this direction, but how quickly they should start and with whom they should do it.

The FileMaker ecosystem — Claris, the consulting firms, the independent developers, and the growing suite of purpose-built tools like FM Dojo — is providing credible answers to that question right now. That’s not hype. From where I sit, that’s a platform earning the next phase of its relevance.