Release notes

Also here: Roadmap · API changelog · MCP changelog · SDK changelog · Glossary · Launch stories on the blog

Release	Shipped	What it was about
5.5	Jul 2026	Code automations in Python, written by you or by an agent · Rollup and rebuilt Formula fields
5.4	Jun – Jul 2026	Bulk work at scale, full-dataset export, ordered attachments, a curated public API
5.3	May – Jun 2026	Inline cell editing, a rebuilt rendering core, descriptions as AI context
5.2	Apr – May 2026	CSV/XML/XLSX imports mapped by AI, MCP tokens for your own AI tools
5.1	Feb – Mar 2026	Dark theme, the reworked filter, a better Make integration flow
5.0	Jan 2026	v5 — modules at the core, AI Mapping, a Remote MCP server in every account
4.11	Sep – Nov 2025	Two-factor authentication, `OR` in the Segmenter, bulk record editing
4.9 – 4.10	Aug – Sep 2025	The work 4.11 announced, plus the rendering layer large systems depend on
4.8	Jul 2025	The Segmenter, progressive loading, manual edits protected by priority

Version 5.5

Shipped 10 – 23 July 2026

Automations you can write in Python — or have an agent write for you. Automations stop being a fixed menu of actions this release: a step can now be your own code, running in an isolated sandbox, and Agent Chat can build that step end to end from a sentence — writing it, testing it against your real data, fixing what fails, and shipping it live.

Code automations — a Code action runs your own Python as an automation step, in a secure isolated sandbox, with credentials held as Secrets rather than written into the script.
Three ways to write that code — by hand in the editor, by describing it in Agent Chat (the agent writes it, wires the trigger, stores the Secrets, tests, and deploys), or from your own agent over MCP, which can create and edit code automations the same way.
Rollup fields — a new field type that aggregates a field across a record's linked records: the total value of a company's deals, how many are open, the latest order date for a customer. On higher plans.
Formula fields, rebuilt — the editor inserts fields as chips instead of raw text and validates live, over an expanded function library: math, logic, text, dates, arrays, conversion, and record metadata. On higher plans.
Richer automation triggers — a scheduled trigger for time-based work, record filters on the Created / Updated / Deleted triggers so an automation only fires for the records it should, and custom fields available in the Record-updated trigger.
Incremental loading for modules — returning to a large module fetches only the records that changed since your last visit instead of the whole table. See Processing large data.
Vector database controls — vectorization is switchable per system, per space, and per field, so you decide exactly what is embedded for semantic retrieval. Rolling out to selected systems.
Webhooks carry what changed — the Update event for module records now includes the changed data, and high-volume webhooks can batch their deliveries. See Actions and triggers.
"Custom modules" are just modules — the terminology is unified across the product, the API, and the MCP server. On the API this is the custom-field → field rename: the old routes stay as deprecated aliases that answer with a Deprecation header and a Link to their successor, so nothing breaks while you migrate. See the API changelog.

Also in this release: move a field group to another view straight from its context menu · modules get a default logo and icon, and land in the menu on creation · invite an administrator while the system is being created (see Inviting users) · the permissions you pick for an API or OAuth token are applied the moment it is created, so a new token works without a second save · Appflows are hidden on new systems, with management of already-installed ones kept.

Agent Chat and the MCP server ship on their own cadence rather than with the release. Over the same weeks, Agent Chat gained web search, attachment previews, and reading of .xlsx and document files, and the MCP server went from v2.2 to v2.3 — automation, menu, status, and label tools.

Version 5.4

Shipped 16 June – 10 July 2026

Scale, and taking your data with you. Bulk work got faster where the systems are biggest, and exports stopped being limited to what happened to be on screen.

Faster bulk operations — bulk create, update, and delete handle high volumes more efficiently, and the front end keeps field-heavy modules quick to open through virtualized field dialogs and a lighter rendering layer.
Full-dataset export — the export dialog now offers Visible fields for what is in your current view, or Full dataset for everything, in a single file, ready for reporting, backups, or a migration.
Ordered attachments — drag files into the order you want and mark one image as the default. That order travels with the data, including through the API, so connected tools receive the files arranged as you intended.
A curated public API — 5.4.0 is the release that defined the public API surface: the endpoints we support, document, and keep stable, drawn as an explicit line. Every release from here is diffable against the last, which is what makes the API changelog possible, and what the SDKs are generated from.
Labels API — full read/write access to labels, so tagging can be automated like anything else. See Labels.
Bulk status changes drive automations — changing status on a selection now fires the trigger for every affected record, not just the first, and the Delete action posts its payload so "record deleted" triggers reliably. See Actions and triggers.

Also in this release: pin a column from its context menu · record detail shows who created a record and when, with a cleaner, editable remote ID · date and date-and-time fields edit inline the same way in the table and in the record detail, on one shared foundation · leaving a screen with unsaved changes prompts first · AI fields appear reliably in lists and offer only connected field groups.

Read the announcement →

Version 5.3

Shipped 19 May – 8 June 2026

A faster core, and agents that understand your data. The rendering and state layer was rebuilt from the foundation up, and modules gained the descriptive context an agent needs to work accurately.

Inline cell editing — click into any cell in a module and edit it in place. No edit mode to toggle, no dialog for a one-character change. See Manage records.
A core that keeps up — tables, record details, and dashboards open faster and stay smooth on the heaviest systems, including those in the hundreds of thousands of records. See Processing large data.
Descriptions as AI context — every module and space has a Description field for notes and metadata, and that description is fed to AI features and connected agents as context.
More dependable MCP — filtering accepts any parameter, including identifiers such as spaceId and values that merely look numeric, and connected agents are pointed at your module schema first so they build on the fields you actually have. See Connect via MCP.
Smarter AI Import — define transformation steps as data arrives, standardizing formats on the way in (unifying date formats, for example), so what lands in your modules is clean from the start. See Import data from a file.
Bulk actions on linked records — the same options, behavior, and permissions you have in your main tables, applied to a relationship.

Also in this release: reorder spaces and menus by drag and drop, in the menu editor · linked-record fields with no connected module are flagged in the list before they can affect your data · setup wizards use the full width of large monitors · the Segmenter confirms before closing with unsaved changes · real-time updates hold their connection on a more secure channel, so the screen stays in sync without a manual refresh · search keeps focus as you type and reaches into grouped fields.

Read the announcement →

Version 5.2

Shipped 15 April – 13 May 2026

Any file in, mapped by AI. Imports stopped being a mapping exercise, and connecting your own AI tooling became a documented, first-class flow.

CSV, XML, and XLSX imports with automatic mapping — upload a file and its structure is mapped into your modules, with columns created where they are missing. See Import data from a file.
AI Mapping, production-ready — several AI mappings in one scenario, editable mapping before it runs, a preview of the queue with rows you can drop, transformations shown separately from the mapping, input validation, and parallel execution.
MCP tokens — connect Claude, ChatGPT, Cursor, and other MCP clients to any module, with a step-by-step walkthrough per client under Connect your AI tool.
Last-changed tracking on records — an updated timestamp that also reflects changes to custom field values, so "what moved since yesterday" is answerable. It is what the incremental loading in 5.5 is built on.
Simpler Make setup — no API token during the welcome flow; it is supplied when you create a specific integration, and if your token spans several organizations you choose which one to use.

Also in this release: the key field is marked in the table header with an icon and a tooltip (see Key column) · resizable side panels per module · attachments move into their own tab in the record detail · faster duplicate detection and queue handling on large imports · organization and team selection in the welcome flow.

Read the announcement →

Version 5.1

Shipped 16 February – 17 March 2026

Settling v5 in. The release that followed the v5 launch was about depth: the filter and import paths hardened, the Make integration reworked, and a dark theme for people who live in the app all day.

Dark theme — a theme selector in Design settings, with the whole interface following it, login screen included.
The reworked filter panel — faster, and built for the compound queries real data needs. See How do I filter records.
A better Make integration flow — clearer mapping, a renamed and reorganized Boost.space app in the Make engine, and that new app used for modules.
The file-import path built — CSV mapping and import blueprints on the API, and AI Import scoped to modules. This is the groundwork the file import announced in 5.2 sits on.
Webhooks improvements — automation actions send full record data, with better operators to act on it, and webhooks became deletable through the API.
Stricter field validation — field names and multi-field creation are validated up front instead of failing on save.
AI usage metered per plan tier — see Plans, tiers, and usage limits and AI pricing.

Also in this release: database query paths sped up · the MCP server gains PKCE and additional token authentication methods for stricter OAuth clients (see the MCP changelog).

Version 5.0

Shipped 15 – 26 January 2026

The database your automations and AI agents were missing. v5 is the release where modules became the product: your own structure, generated by AI if you want it, readable and writable by your automations and by any MCP-compatible agent.

A database behind your Make.com scenarios — connect Make with an API token and your scenarios read from and write to a shared data layer instead of passing records into the void. Two-way sync, full visibility of what arrives from which scenario, priority rules for when two scenarios touch the same record, and a rebuilt Boost.space app in the Make engine.
AI Mapping — drop an entire data bundle from any app into Boost.space and AI reads it, recognizes each field, and maps it to the right column, creating the column if it does not exist.
A Remote MCP server in every account — connect Claude, ChatGPT, Cursor, Windsurf, or any MCP-compatible client over OAuth or an API token, and your agent can read and act on your business data through a proper protocol. See Connect via MCP.
Modules at the center — describe what you want to track and AI builds the module, field groups, relationships and types included. Standardized fields (email, phone, address) keep your own structure compatible with platform-wide automation. Modules appear under their real names everywhere — filters, webhooks, settings.
AI throughout the data — an AI toolbar on every table: AI fields that generate values, data transformation for reshaping values across records, and data validation for catching inconsistencies before they spread. Plus an in-app AI chatbot for questions about Boost.space itself.
Button fields — a clickable button in the row that fires a webhook for that specific record. One click, one automation.
Linked records as tables — flip a toggle and a relationship renders as a full interactive table inside the record.
Inline record creation — click + in the table, start typing, done. See Manage records.
A reworked filter — the old sidebar filter is gone, replaced by something faster and built for compound queries.
Color-coded select options — assign colors to single- and multi-select values for instant visual grouping.
Social login — sign in with Google, LinkedIn, or X.
Simplified settings — advanced options behind toggles, off by default. Clean for new systems, complete for the ones that need it. See System settings.

⚠ Breaking change — manual edit priority moved from High to Medium. Manual edits and scenario updates now share the same default priority, so a manual edit no longer automatically wins. If you relied on that, review your priority settings — see API token prioritization.

In the same weeks, the Make Integrator that powers scenarios shipped its own large update, which arrived in Boost.space automatically: a refreshed interface, a Gmail app that no longer needs a Google Cloud project for personal accounts, a rebuilt HTTP app with built-in pagination, reusable data structures and a secure keychain, and AI web search inside a scenario.

Read the announcement → · Integrator update →

Version 4.11

Shipped 23 September – 27 November 2025

Security and data control. Account protection arrived, and the Segmenter turned into a proper query tool.

Two-factor authentication — enable 2FA with an authenticator app from your profile, and every login asks for a code.
OR logic in the Segmenter — segments on several alternative criteria, so "either of these" is one segment instead of two.
Bulk record editing — select records in any table, choose Bulk edit, and change fields across all of them at once.
Segmenter management — the matching record count shown before you save a segment, drag-and-drop ordering, and a central per-module overview where segments can be edited or deleted.
Segments for modules — the same filtered working sets on your own modules, not only the built-in ones.
Persistent column widths and a loading indicator — column widths survive a refresh, and table loads report their progress as a percentage.
Modern date pickers — redesigned, with the calendar week starting on the correct day for the interface language.
Rich text, properly — images and tables inside the rich-text editor.
Colored options for select fields — stored on the field itself, ready for the UI that landed in 5.0.
WEBP attachments — uploaded and previewed correctly.

Also in this release: bulk endpoints for modules in the API · a Table view for scenarios with customizable scenario properties · the MCP server reached its first release candidate in November 2025, on the way to the stable line that shipped with 5.0 (see the MCP changelog).

Read the announcement →

Version 4.9 – 4.10

Shipped 6 August – 10 September 2025

Building what 4.11 announced. 2FA, the Segmenter's OR filter, and bulk editing were all written across these two releases and announced together in 4.11 — the normal shape of things, and why the two share one entry here rather than each being given a headline. Alongside them, the rendering work the largest systems depend on.

Aggregations behind the list endpoints — a count or a total comes back as one number instead of a page of records. The MCP server picked this up the same month, in its v0.4 line, which is why a connected agent can answer "how many" without reading everything.
OR in filters, server-side — the parameter under the Segmenter's OR option.
Bulk edit in the API — the endpoint behind the interface that arrived in 4.11.
Virtualized dropdowns — long option lists render only what is on screen, and virtualization became switchable per component. Tables had been virtualized in 4.8; this is the rest of the interface catching up.
Loading skeletons — tables and detail views show their shape while data arrives instead of a blank screen.
React 19 — the interface moved to the current major, unlocking the concurrent rendering the performance work above depends on.
Scenario input and output parameters — a scenario declares what it takes and what it returns, instead of that being implicit in its steps.
Steadier AI features — fixes to the AI scenario generator and to data validation, plus custom field tooltips so a long field name is readable where it is used.

Version 4.8

Shipped 9 – 30 July 2025

Large datasets became workable. The Segmenter, virtualization, and progressive loading arrived together — the release where "too many records" stopped being an answer.

The Segmenter — a saved, filtered working set available in every module, so a table shows the records you care about instead of everything. Activated automatically on systems past 100,000 records.
Status Center — connection state, system version, and browser details in the left panel, with sync problems named rather than guessed at.
Manual edits protected by priority — data edited by hand is written at high priority, so only the right API tokens can overwrite it.
Progressive loading — custom field groups and table columns load as you reach them, keeping field-heavy systems responsive.
256 MB file uploads — the file field limit raised, so contracts, design files, and full exports fit as attachments.
Faster record loading — record details open quicker, and returning to a table no longer triggers a reload you did not ask for.
Clear filter in one click — remove every active filter in a table at once.
Expanded context menus — act on a module, an appflow, or a custom field directly from its context menu, in settings and in detail views, instead of opening a form to get there.
Email verification before the Integrator — an unverified account is sent to verify before it can create scenarios.

Read the announcement →

Earlier releases

Boost.space release notes go back to 2.0.30, in October 2021, and everything before 4.8 lives on the blog:

All release notes on the blog →

Two other trails cover the same ground for developers: the API changelog from 5.4.0 on, and the MCP changelog from the server's first version in June 2025.

Roadmap

API changelog