FileMaker 2026 Executive Summary

Introduction

Before we get into the release itself: in May 2026, Claris CEO Ryan McCann published How Claris is building for what comes next, which lays out where the platform is heading. It’s a strategically important read, and it sets expectations you’ll want to calibrate carefully against what 26.0.1 actually ships.

The headline of McCann’s post is that Claris is leaning into AI-assisted, agentic development on top of FileMaker — making FileMaker a first-class deployment target inside agentic coding tools like Claude Code, Cursor, and Codex, with internal work already underway on conversational web-native UI generation and on agents that read your schema and write native FileMaker schema (scripts, calcs, tables, fields, value lists, custom menus…). Developer previews of those capabilities are scheduled for “later this summer, following the release of FileMaker 2026.”

That last clause is the one to internalize: none of the agentic coding capabilities described in that blog post are in 26.0.1. What McCann explicitly attributes to this release is the resiliency-and-infrastructure work — Standby Server and Remote Backup — that, in his framing, “lays the groundwork for agentic development.” Those features are real, they are in 26.0.1, and we cover them below. The Claude Code / Cursor / Codex story, the conversational UI generation, and the script-writing agents are all coming, but they’re not in this release. If you’ve read McCann’s post and walked away expecting agentic features to land with this upgrade, recalibrate before you brief your stakeholders.

With that out of the way, Claris released FileMaker 2026 — version 26.0.1 — in June 2026. As I’ve done for the last few years, I’m dropping the marketing name on the second mention, and I’m going to refer to this release by its actual version number. Marketing names span multiple point releases (FileMaker 2025 covered everything from 22.0.1 to 22.0.6, and they all behaved differently). When you go to troubleshoot something a year from now, “26.0.1” tells you what you need to know; “FileMaker 2026” does not. Many of us at Soliant participated in the ETS (early access) program for this release, so what follows is our curated take on what matters for developers and the people who make the decisions to foot the bill for developing on the Claris FileMaker platform and acquire the licenses for it. This is not a comprehensive feature list — for that, read the release notes for FileMaker Pro, FileMaker Server, and FileMaker Go. What follows is what we think is worth your attention, with our reasoning and the gotchas we’ve found along the way.

Platform and OS support

The compatibility envelope tightens again this year. The headline points:

• FileMaker Pro 26.0.1 runs on macOS 15 Sequoia or later and Windows 11. Windows 10, macOS Ventura, and macOS Sonoma are no longer supported.
• FileMaker Server 26.0.1 runs on Windows Server 2022/2025, macOS 15 or later, and Ubuntu 22/24 LTS. Server 2019 is dropped, and macOS 14 is dropped.
• FileMaker Go 26.0.1 requires iOS/iPadOS 18 or later. Building with the iOS App SDK requires Xcode 26.

Connecting clients to a host: hosts running FileMaker Server 2024 (version 21) or earlier are not supported by 26.0.1 clients — upgrade your servers first or you’ll be stuck. A subtle one worth flagging out loud: this is likely one of the last FileMaker Pro/Server releases to support Intel Apple hardware on macOS. Apple is widely expected to drop Intel support past macOS 26, and FileMaker has historically followed Apple’s curve. If you have Intel Macs in production today, factor that into your hardware refresh plan now, not in 18 months when it becomes a fire drill.

Headline Features

Persistent data store

This is the feature I expect to change how we build things, and we’ll publish a deeper Soliant blog post on it.

The persistent data store is a brand new concept in FileMaker. It is schema, not data — a set of named values saved as part of the file’s schema, addressable from any script or calculation, in any context, without record navigation, without table occurrences, without globals’ baggage. The new script step Configure Persistent Data writes entries; the new functions GetPersistentData ( name ; instanceID ) and ListPersistentDataIDs ( name ) read them back.

Why it matters: every solution of any size has needed a place to stash a small amount of file-wide state — the version number of the app, configuration for an add-on instance, JavaScript library code for a web viewer, the prompt template you send to an LLM, environment-specific endpoints. Today, we cobble this together with a settings table, globals, custom functions returning hard-coded values, or external-files-as-config. The persistent data store gives us a first-class place to put it.

A few things to internalize before you reach for it:

• Persistent data persists with the schema, so it travels with Save a Copy as XML, with the upgrade tool, and with add-ons. Add-on developers in particular should be looking at this as the proper home for per-instance configuration.
• Performance trade-off: variables live in the client’s memory, persistent data lives on disk in the file (and may not even be local if the file is hosted). It also synchronizes across clients. The blunt summary from the Claris team during the ETS program: “it will be slower than variables.” Don’t reach for the persistent data store as a drop-in replacement for $variables — use it for values that you set once (or rarely) and read across many contexts. Reading and writing it inside a tight loop, or anywhere a $variable would have done the job, is the wrong tool.
• Values are cached on the client after first read, so the second access in a session won’t go back to the server unless another session has changed the value — same model as record data in a hosted file.
• Each entry is named and supports multiple “instance IDs,” which is the mechanism for storing per-add-on-instance configuration in the same name.
• Worth repeating from the first bullet: because persistent data lives in the schema, its values are visible in Save a Copy as XML output. Don’t store API keys, OAuth secrets, passwords, or anything else sensitive in there. Treat it as configuration, not as a vault.

This is the kind of feature that doesn’t look big in a release-notes bullet but quietly changes how you architect things. We’ll write more on this separately.

PDF script steps

FileMaker has been able to save records as PDF for as long as anyone can remember, and we’ve all needed to combine PDFs from multiple sources at one point or another — usually by shelling out to a plug-in or building a workflow on the server side. 26.0.1 adds a proper script-step API for PDF assembly:

• Create PDF — start an empty PDF in memory.
• Open PDF — open an existing PDF in memory.
• Append PDF — add pages to the open PDF.
• Close PDF — save the open PDF to a file path, container, or variable.
• Cancel PDF — discard the open PDF without saving.
• Print PDF — print a PDF from a path, container, or variable.

The existing Save Records as PDF step now has a Save to option that lets it append to the currently open PDF, which is the bridge between the layout-based PDF generation we already had and the new assembly model.

Why we like it: composing a finished document from a cover page, a layout-rendered detail section, an externally provided attachment, and a generated summary is a workflow we’ve built dozens of times for clients, and it always involved either a plug-in or an Insert from URL round-trip to a microservice. Now it’s native. It’s also supported server-side, which is where most of these workflows belong anyway.

That said, native doesn’t always mean best. PDF assembly at volume is CPU- and memory-hungry, and pushing it into a dedicated microservice is still a perfectly valid pattern when performance matters — it gets the load off the FileMaker Server box (or off the user’s client) and lets you scale the PDF workload independently. Reach for the native steps for the everyday case; reach for an external service when you’re building documents at scale or on a hot path where every millisecond on the FMSE counts.

A reminder you’ll need: the PDF-Writer library was updated to 4.8.1 in this release on both Pro and Server. Anywhere you generate PDFs — these new steps, Save Records as PDF, the Print script step on Server — needs to be regression tested. Glyphs, fonts, page breaks, and embedded images are all worth a careful look.

Standby Server (high availability)

This one we will do a Soliant blog post on, because it’s the biggest server-side architectural change in years.

FileMaker Server 26.0.1 ships a Standby Server configuration: a second server running in lockstep with the primary, kept in sync via incremental backup, ready to take over as the primary when you trigger a switchover. This is the high-availability story we’ve been writing custom for clients with NFS mounts, snapshots, and DNS tricks for years, and it’s now a first-class feature of the platform.

What you need to set one up:

• Two machines on the same general OS — macOS to macOS, Windows to Windows, Ubuntu to Ubuntu. No mixing.
• An SSL certificate strategy that covers both hostnames — a SAN cert, a wildcard cert, or two single-host certs (one per machine) all work.
• OpenSSH 9.0 or higher.
• A full FileMaker Server install on both machines (the standby isn’t a stripped-down agent).
• The incremental-backup interval and folder configured on the primary, since that’s the synchronization mechanism.

Things copied to the standby on switchover include plug-ins, session timeout values, database filtering, HTTPS tunneling, Nginx (Windows), Admin API PKI, external authentication settings, schedules, notification settings, and external container storage. You can also create the standby from a hypervisor snapshot rather than pairing over SSH, which opens this up for AWS-style fleets.

What it is not: this is not synchronous replication, and it is not automatic failover. You initiate the switchover yourself when you decide the primary is unhealthy, and any window of activity between the last incremental backup sync and the switchover is at risk. For most small-to-medium deployments, that trade-off is the right one — the cost and complexity of true synchronous HA is rarely justified for FileMaker workloads. But know what you’re buying.

Multiple Script Engine processes

FileMaker Server has used a single FileMaker Script Engine (fmsase) process for as long as I’ve worked with it, with multi-threading inside that process. In 26.0.1 you can now spawn multiple fmsase processes simultaneously — three, by default, with a configurable number of cores per secondary engine.

The configuration lives in ClarisConfig.json:

"MultipleSASE" : true,
"CoresPerSecondarySASE" : 4

CoresPerSecondarySASE can be set anywhere from 4 to 16. The default of three secondary engines means that a value of 4 cores per secondary requires at least 12 cores on the box.

Why we like it: three different buckets of functionality have always shared a single fmsase process — Perform Script on Server (PSoS), scheduled scripts, and the OData API — and on a busy server with a lot of long-running scripted work this becomes a bottleneck. One badly-behaved script (or a heavy OData consumer) can starve everyone else. Splitting work across multiple engine processes provides real parallelism, not just thread-level concurrency, and the OS scheduler can move them across cores independently.

The honest question we still have: how this stacks up against the existing multi-threading inside a single fmsase is something we’ll need to benchmark carefully. We’d expect the biggest wins on workloads that are CPU-bound (the inside of a single fmsase was always somewhat constrained) rather than I/O-bound. If you have a script-heavy server, and the box has the processing capacity (number of cores) to support it, this is a knob to know about.

Remote Backup

A third headline server feature: the new Remote Backup functionality backs up databases from the secure-storage and default database folders to a remote drive, configured through the Admin Console (with Claris Customer Console involvement). For shops that have been hand-rolling rsync, robocopy, or restic-based off-box backups, this is a nice consolidation into the supported product surface.

A known issue worth noting in the initial build: the connection-test UI may report a failed connection even when the connection is fine — refresh the page to get the correct status.

A note on disaster recovery and business continuity: Standby Server and Remote Backup are welcome additions to the platform. We’ve talked about DR and BC at many DevCons and Engages and in many blog posts over the years — it’s something we’ve spearheaded in the FileMaker community for a long time, and it’s in our DNA. For instance, in our own soliant.cloud managed hosting environment we offer SnapBacks (snapshot-based backups that eliminate paused-database time and reduce data-loss exposure), OptiFlex DR (a fully functional copy of your entire FileMaker Server stack in the cloud, ready to launch at a moment’s notice), and on-prem-to-cloud standby server configurations for clients who keep their primary on-premises but want a hot cloud failover. If you want defense-in-depth around your FileMaker workloads — beyond what 26.0.1 ships out of the box — that’s the layer we operate at.

AI Feature Refinement

If 22.0.1 was the year of “AI is here, and you’re going to use it,” 26.0.1 is the year of “now let’s actually make it usable in production.” There’s no single big new AI feature; instead, there’s a long list of refinements that, taken together, materially change what’s practical. They split cleanly into two buckets: improvements you see from inside your FileMaker solution (Pro, Go, WebDirect), and improvements on the Claris AI Model Server itself.

From inside your FileMaker solution

Image captioning

Two new script steps — Insert Image Caption and Insert Image Captions in Found Set — send container-stored images to an image-captioning model and return a generated description. This is a third model type alongside text generation and embeddings.

The Claris-tested captioning models are Salesforce/blip2-opt-2.7b (added by default on a clean install of the model server) and Salesforce/blip-image-captioning-large (which produces some unexpected words in our testing). The list will grow. (See our “The Claris AI Model Server and AI-Generated Captions for Images Stored in FileMaker” post)

Field annotations for AI

Until now, if you wanted to give an LLM context about a field, you had to put [LLM] in the field comment — which polluted the comment for human readers. 26.0.1 adds a proper field annotation in the Advanced Options for Field dialog that goes into the Data Definition Language (DDL) sent to the model and stays out of the field comment. The new FieldAnnotation() function reads it back.

The behavior to remember: when no field in a table is annotated, the DDL includes all fields. When any field is annotated, the DDL includes only the annotated fields. The old [LLM] field-comment tag is still respected for backwards compatibility, but the new annotation wins when both are present. This is actually the right design — it gives you fine-grained control over what schema you expose to a model — but it’s a behavior you need to internalize before you start sprinkling annotations onto a production solution.

The annotations also flow through OData metadata, which is useful for any tooling outside FileMaker that wants AI context.

Google Gemini, Cohere image embeddings, more knobs

The Configure AI Account script step now supports Google Gemini as a commercial provider for both text generation and embeddings. Cohere now supports image embeddings through Insert Embedding, Insert Embedding in Found Set, and GetEmbedding. Insert Embedding and Perform Semantic Find both gained a Parameters option for provider-specific settings (e.g. dimension).

The Parameters option on Generate Response from Model, Perform SQL Query by Natural Language, Perform Find by Natural Language, and Perform RAG Action all support CURLOPT_TIMEOUT so you can finally control how long a slow model is allowed to think before you give up on it.

We continue to recommend running your own model server, with the same security and privacy reasoning as last year, but the bench of viable commercial providers keeps growing for the cases where it’s the right call.

RAG script-step improvements

The Perform RAG Action script step is meaningfully better in 26.0.1:

• It now returns the document ID assigned when you add a text or PDF document, so you can store it for later removal without parsing GetRAGSpaceInfo output.
• Per-request Similarity Threshold and Number of Top-Ranked Results can be set in the Parameters option, overriding the global defaults from Admin Console for that one call.
• Tokens per Text Chunk is configurable per-add-data call, so you can use different chunk sizes for different content types or languages without wrestling with a single global value.
• JSONL is supported as an input format, which is the right shape for chunking FileMaker table data — each line becomes its own chunk.
• Vertical text detection is available when adding container data containing scanned documents (think Asian-language scanned forms).

Insert from URL: a small win that punches above its weight

When Insert from URL retrieves a response with MIME type application/json into a variable, the JSON is now automatically parsed and cached — equivalent to running JSONParse() on it. Subsequent JSON operations on that variable are dramatically faster. This is a quiet performance improvement that will compound across any solution making heavy use of REST APIs, AI calls included. (See our “Changes to Insert from URL Script Step in FileMaker 2026” post)

On the Claris AI Model Server

Independent lifecycle and concurrency

The AI Model Server now gets its own version number (starting at 26.0.1.1), can be upgraded independently of the main FMS via Admin Console, and ships a separate installer option for AI-only deployments. Multiple Python processes mean embedding and text-generation workloads run concurrently rather than fighting each other for a single Python interpreter.

Embedding model type unification (an upgrade gotcha)

The server side has been refactored so that text-embedding and image-embedding models are now one “embedding model” type. You’ll need to remove and re-add your existing embedding models after upgrading to register them under the new type. This is the kind of operational gotcha that bites you in production if you’re not paying attention — plan for it in your upgrade window.

RAG: faster, with a breaking change

RAG embedding generation is now batched and noticeably faster on large content sets. RAG space IDs are now UUIDs instead of incremental integers — a small breaking change worth knowing about if you wrote anything that parses them.

Wider GPU and OS support

AMD ROCm GPUs are supported on Ubuntu (in addition to NVIDIA), and the vLLM engine is available on Ubuntu/CUDA setups for serious throughput. CUDA is bumped to 13.0. The macOS AI Model Server now requires Apple Silicon — Intel Macs are not supported for the model server. The environment manager has switched from Conda to Mamba, which is mostly an internal change but may affect upgrades from older installs.

gpt-oss and the prompt-engineering reality

The model server now supports gpt-oss (in 8-bit and MXFP4 quantizations on Apple Silicon, and the standard 20b version), as well as EmbeddingGemma and a wider set of multilingual embedding models. From our testing, gpt-oss requires highly customized instructions to use tools and return acceptable responses — the Claris team has shared an example prompt structure (generated by Claude 4.5, of all things) that worked in their testing. The take-away: the difference between “this open-source model is unusable” and “this open-source model is great” is often a careful system prompt, and that’s a skill the FileMaker community is still building.

FileMaker Pro: developer quality of life

A long list of smaller features that, individually, are minor — collectively, they represent a meaningful upgrade in day-to-day developer ergonomics.

Window UUIDs

Every window now has a UUID, accessible via the new Get(WindowUUID) function, and Select Window, Close Window, Move/Resize Window, and Set Window Title all accept UUIDs as identifiers. This is the right fix for the long-standing problem of two windows sharing the same name (or one window’s name changing out from under you). We’ve publish a Soliant blog post showing this in action — the most compelling use case we’ve found is a multi-window back-button stack combined with GetRecordIDsFromFoundSet(). (See our “Managing Windows by UUID in FileMaker” post)

GetRecordIDsFromFoundSet() — now relationship-aware

The function picks up an optional second parameter: a table-occurrence name or a portal object name. With it, you get the IDs of the related records (single-hop only) or, for a portal name, the IDs of the visible portal rows, including the effects of portal filtering and sorting. Useful for picker workflows where the user selects multiple related items and you need to capture their keys for later record creation. We have a Soliant blog post on this.

Smart Layout Inspector (macOS)

The macOS Inspector pane has been redesigned: two tabs (Appearance, Data) instead of four, and only the controls relevant to the selected object are shown. Less visual noise, much easier to teach. Many of us have years of muscle memory from the old Inspector panes, so adjusting to the new one will cause some friction before we settle into what we like about it.

Customize field display names

Fields can now have a custom display name for use in the Sort Records dialog, the Specify Field Order for Export dialog, exported Excel files, and the Table View column header, set in the Advanced Options for Field dialog. The mechanism is a JSON object with the keys fm_common, fm_sort, fm_export, and fm_table_view. Combined with the new FieldDisplayNames() function, you can finally show users human-friendly names without changing the underlying field names. Long-overdue.

Configure field entry by calculation

The “Allow entry into field” layout option is now a calculation. Return 1 for full editing, 2 for read-only-but-selectable (good for copy/paste of values), or any other value for view-only. A real improvement, but not the full read-only story we’ve been wanting — the field still shows no scrollbar in view-only mode, so users can’t scroll to read content that overflows the field. Half a step forward.

Script Workspace: collapsible comments and disabled steps

Long comment blocks and disabled (commented-out) script steps can now be collapsed in the Script Workspace, alongside the existing collapsing for control structures. If you write — or read — long scripts, this is a meaningful change. Combined with the existing collapse for If, Loop, and Open Transaction blocks, the Script Workspace is finally as navigable as a modern code editor.

SQL gets foreign keys

FileMaker SQL now supports FOREIGN KEY syntax in CREATE TABLE and ALTER TABLE, meaning you can create relationships from SQL — including relationships specified by an LLM in the natural-language SQL workflow. A clean, predictable extension of what was already there.

A handful of others worth knowing about

• Flush Web Viewer Cookies — a new script step. If you have a kiosk-style or shared-device deployment where one user’s web viewer session shouldn’t leak to the next, this is the proper way to clear it.
• Get(GuidedAccessState) — detects whether iOS Guided Access is on, useful for kiosk-mode FileMaker Go deployments.
• Get(AccountPasswordDaysRemaining) — surface password expiration before users get locked out.
• BaseTableComment() — read base-table comments (table comments were new in 22.0).
• GetTextFromPDF on macOS now supports image-based / scanned PDFs, falling back to OCR when the PDF has no embedded text.
• WebP image support in container fields — insert, display, and work with WebP without conversion.
• Re-Login can now target a specific external data source rather than re-authenticating everything at once. Multifile solutions appreciate this one.
• Show Custom Dialog lets you specify dialog size and position (also in WebDirect), and shows a scrollbar when content overflows.
• Manage default fields now has a per-file File Options setting and an application-wide setting for whether default fields are added to new tables. This setting only applies to new files — there’s no in-place opt-out.
• Sort blank values last when sorting ascending — a tiny QoL win that has annoyed users for two decades.
• Replace Field Contents with auto-enter skipped — Full Access only. Useful for one-off bulk updates where you don’t want modification timestamps or other auto-enter calcs to fire.
• Two new privilege options introduced in 22.0.4 (Manage database/data sources/containers/custom functions and Manage custom menus) are now the gate for Data Viewer access, separating script debugging from Full Access.
• ProAnalytics (“phone home”) is new. FileMaker Pro now records and sends product analytics to support future feature plans — OS, language, version, and various script-step usage. We’re still gathering details on what’s sent, how often, whether it can be logged, what happens when blocked at the network layer, and what the opt-out path is. Before you roll 26.0.1 out broadly to a regulated environment, get clarity on this from Claris and confirm it meets your data-handling policies.

FileMaker Server: beyond the headline features

Beyond Standby Server, multi-FMSE, and Remote Backup, the server side picks up a long list of operational improvements.

Crash recovery for connector services

FMS now automatically restarts these services if they quit unexpectedly:

• FileMaker Data API (fmwipd)
• OData (fmodata)
• Web Publishing Engine (fmscwpc)

This is exactly the right level of self-healing — no more 3 a.m. pages because OData fell over and stayed down. Note that this is restart, not a deeper “monitor and alert” story. You should still be watching for repeated restarts, because that’s a signal that something is wrong upstream.

Admin API growth

The Admin API picks up endpoints for managing the AI Model Server (start/stop, models, all settings) and managing Standby Server configuration (server role, allowed standby host, standby list, pairing, SSH setup, incremental backup, switchover). For shops automating FMS deployments with infrastructure-as-code, this is the API surface you’ve been waiting for.

OData

OData’s metadata gets richer in 26.0.1:

• Field comments and AI annotations are now in the metadata.
• Field options and scripts are in the metadata.
• Scripts can now be executed by script ID with a URL of the form /odata/v4/<db>/Script.FMSID:<id>.
• Webhooks support a new maxFailedAttempts field on subscription, log notification errors to fmodata.log, and return a ROWID of -1 when all records in the subscribed table are deleted or the table is truncated (so you don’t miss those events).

We’ve published a dedicated blog post on the OData changes — see our “Enhances to OData API in FileMaker 2026” post.

Authentication

• Custom OAuth can now specify a return URL pointing to an external website, not just FileMaker Server. This is exactly what you need to wire OAuth flows into a web app that uses its own backend for authentication.
• PKCE is supported for OAuth (OAuth 2.1), required by some identity providers.
• A handful of dependency CVEs (brace-expansion, on-headers, send, cookie, body-parser, @babel/runtime) are addressed in Admin Console.

Installation and deployment

• Installation on macOS Ventura is blocked — match your OS plan to your FMS plan.
• Uninstalling FMS on macOS and Ubuntu now preserves encryption keys, cached RAG data, and downloaded AI models (matching Windows behavior). This is a real operational win for in-place upgrades.
• macOS: the installer now offers a CLI-tools-only install option, also available via Assisted Install (Deployment Options=4). This means that you can install the FMDeveloperTool, FMUpgradeTool, and FMDataMigration CLIs without having to also install the rest of FileMaker Server.
• Ubuntu: the Assisted Install file supports a Preserve Firewall flag for fresh installs (Preserve Firewall=1); existing UFW rules are now preserved automatically on upgrade. Existing users of our unified Assisted Install repo (read the blog post) — we’ll update for these new options.
• Ubuntu: the installer can now run FMS as a non-default OS user, matching macOS and Windows.
• Windows: the installer can set NtfsDisable8dot3NameCreation via the Assisted Install file (Disable 8dot3 Name Creation=1) for performance.

Persistent IDs in containers and clusters

Get(PersistentID) on FileMaker Server now returns a stable identifier across server restarts, reboots, and upgrades, derived from an OS-level machine identifier at install time and stored in the server’s configuration. Most importantly, this fixes the long-standing problem where Docker containers using bridge networking would get a new MAC address (and thus a new persistent ID) on every restart, breaking saved encryption keys and SMTP configurations. If you’ve been running FMS in containers, this alone is worth the upgrade.

Server-side script step support

The Export Field Contents script step is now supported in scripts run by FileMaker Server, the FileMaker Data API, and the OData API. Worth knowing: there’s been a long list of “this works on the client but not on the server” gotchas for this script step over the years.

WebDirect

WebDirect picks up some real maturity in 26.0.1:

• JDK 21 is the new minimum (up from JDK 17).
• Section 508 / WCAG accessibility improvements, including a new Accessible Keyboard Navigation option in Web Publishing > Connectors. With ARIA-compliant keyboard controls enabled, a blue frame highlights the current selection and users interact via Enter/Space.
• Paste operations are handled client-side instead of round-tripping to the server — perceptibly faster.
• The idle-timeout warning duration is now configurable (Admin Console > Configuration > FileMaker Clients > “Timeout dialog duration for FileMaker WebDirect”). The previous fixed 30-second warning was both arbitrary and, for some workflows, too short.
• The Action is Running dialog no longer pops up when a layout redraws because of a window resize. A small thing, but it was deeply annoying.
• Custom field display names are honored in Export Records.

A handful of long-standing rendering and input issues are also fixed — the Japanese-IME text-loss bug, blank charts, the calendar drop-down on mobile, web-viewer flashes, and more. WebDirect is the closest it has been to “boring and reliable” in years.

FileMaker Go

Go 26.0.1 is mostly platform-upkeep:

• iOS/iPadOS 18 minimum.
• iOS App SDK requires Xcode 26 minimum.
• OpenSSL bumped to 3.5.5 LTS.

The notable bug fixes: the Set Web Viewer script step no longer leaves the web viewer blank, the Open URL step no longer mangles # to %23, PNG transparency is preserved when inserting from Photos, and the Japanese calendar format is honored in drop-down calendars. If you have any of these in production today, the upgrade is uncomplicated.

The AI script steps and the persistent data store all work on Go, by virtue of being part of the file’s schema and engine — there’s nothing special you need to do to enable them in mobile workflows.

Tooling: Developer Tool, Upgrade Tool, Save as XML

A trio of tooling improvements that together represent a real step forward for solution lifecycle management.

FileMaker Developer Tool

• New --changeAccountPassword command lets administrators reset any database account password from the CLI by authenticating with admin credentials, no need to know the existing password.
• --saveAsXML aligns with the new Save as XML dialog: select catalogs to include with -cl, split each catalog into its own XML file with -sc, include analysis-tool details with -id, and control whether layout-object binary data is stored inline or once in the library catalog with -s.
• --sortBySize now reports table IDs that match TopCallStats.log entries, finally making it possible to correlate file-size analysis with execution profiling.

FileMaker Upgrade Tool (the artist formerly known as the Solution Upgrade Tool)

• New —generateDBFile command regenerates a .fmp12 file from a Save-as-XML output — effectively a re-hydrate / deserialize step, with optional plugin_folder for files that reference third-party plug-ins. Note that this generates new IDs and UUIDs, so it’s not a perfect round-trip, and there are a number of things known not to round-trip in this initial version (read the release notes carefully).
• ReplaceAction gains support for Page Setup, TableOccurrenceNotes, and group objects (groups, portals, slide controls, tab controls, button bars, popover buttons). Patch-based deployment of UI changes finally covers the cases that matter.
• DeleteAction supports removing TableOccurrenceNote entries, by ItemReference for a specific note or by table-occurrence target for all of them.

Things to be aware of

A short list of things that aren’t headline features but will trip you up if you don’t think about them:

• Add-ons created with this version include a Creation_BuildID, and Pro 26.0.1 won’t display add-ons created with a newer version of Pro. If you’re an add-on author and you need to ship an add-on built with a newer Pro to clients still on the older version, you can manually remove Creation_BuildID from the add-on’s info.json. This compatibility check exists for good reasons; it’s not something to disable casually.
• Embedding model registration broke during upgrade — the AI Model Server consolidated text and image embedding into a single “embedding model” type, and existing registered embedding models need to be removed and re-added under the new type after upgrade. Test this in staging.
• The Default Fields setting only affects new files, so if you want to retroactively turn it off in an existing file, you’ll need to clone and migrate.
• PDF generation regression-testing matters this release — PDF-Writer was bumped to 4.8.1 on both Pro and Server.

Conclusion

26.0.1 is not a flashy release on the surface, but the things it ships are the kinds of features that change how we build. The persistent data store gives us a proper place for solution-level configuration. PDF script steps replace a pile of integration work. Standby Server, multi-FMSE, and Remote Backup move FileMaker’s server-side story closer to the operational expectations of the rest of the modern infrastructure stack. The AI improvements, taken together, are the difference between AI-as-demo and AI-as-production. And the tooling improvements quietly make solution lifecycle management considerably less painful.

The OS-support tightening — and especially the looming end of Intel Mac support — is the operational issue worth surfacing to your stakeholders sooner rather than later.

If you’d like help thinking through what 26.0.1 means for your environment, or if you want to dig into any of these features in depth, please get in touch with our team or email me directly at wdecorte@soliantconsulting.com. And keep an eye on the Soliant blog over the next few weeks; the deeper coverage on persistent data store, Standby Server, Window UUIDs, the GetRecordIDsFromFoundSet() update, OData changes, the Insert from URL changes, the SaXML rework, and the Upgrade Tool’s new capabilities will all roll out there.

Release notes

For the complete, authoritative feature lists — this summary is curated, not exhaustive — see Claris’s release notes:

• FileMaker Pro 26.0.1 release notes
• FileMaker Server 26.0.1 release notes
• FileMaker Go 26.0.1 release notes