Momen 3.0 | Released September 2026
This release reorganizes the editor architecture, rebuilds external API integration and the data model editor, adds AI Copilot (an AI builder inside the editor), object types and developer access, and rolls out the new design system throughout.
Product architecture and design system
Editor architecture rebuilt
The old feature-by-feature menu is replaced by four sections — Design, Data, Action and Settings — matching the four stages of building software: interface, database, backend services, and operations. You no longer need to remember which menu a setting hides in: find the right area by common sense, then hit Publish when you are done. Read the docs
| Section | Stage of development | What lives here |
|---|---|---|
| Design | User interface | Pages, modals, components |
| Data | Database, type and variable declarations | Data model, database, objects, enums, resources, client variables |
| Action | Backend services | Actionflow, API, AI, verification code, sign in, payment |
| Settings | Operations | General, permissions, domain, SEO, secrets, script injection |
Design system update
The new design system is applied throughout: the interface of all four editor sections, the full set of design components such as modals and forms, page states such as lists and empty states, plus the icon set and design tokens.
AI Copilot
AI Copilot is now generally available — an AI builder inside the editor that can build pages, components, data models, actionflows, APIs, AI agents and more for you. Read the docs
- Where to find it: top bar → AI Copilot.
- What it can build: covers all four sections — Design, Data, Action and Settings — and can produce a complete front-end and backend configuration. Its output is equivalent to building by hand.
- Usage and billing: Copilot consumption is metered in AI Points. Check your available balance and every transaction under User center → Wallet.
- Grant sources: signup bonus, weekly allowance, purchase, promotion, education grant and support adjustment.
- Metering: consumption is recorded separately for agent turns, context compaction, subagents, skill generation, replays and session titles.
- Points can be drawn from a project’s quota.
Design
- Component system rebuilt: the whole component system was rewritten from the ground up, and the new components come closer to hand-written front-end code in both styling range and rendering performance.
- Custom modals rebuilt: modals are now global elements. A custom modal you create can be opened from multiple places, so the same modal no longer has to be rebuilt on every page.
- Missing style properties filled in: system components gained a batch of style properties they were previously missing, reducing the cases where a design could be drawn but not built. Read the docs
- Reconnecting component styles to the primary breakpoint: component styles that are disconnected from the primary breakpoint are now shown explicitly, and can be reconnected. Read the docs
- Placeholder image for video upload: the video upload component can show a placeholder while an upload runs, instead of an empty area.
Data
Data model
-
Editing interface rebuilt:
- Table details consolidated into five tabs: Field, Constraint, Permission, Vector storage and Trigger, so everything about one table is viewed and edited in one place. Read the docs
- Relationship view: see how tables relate to each other as a graph, which keeps the data structure legible once the table count grows.
-
Field editing improved:
- API name: set a field’s API name when you create it, and see it on the field details.
- Function fields: a new field type with a configurable return value and return type, letting the database compute values derived from other fields instead of repeating the same formula everywhere it is used.
- Continue to create: after adding one field, go straight into the next without reopening the dialog.
- Drag-to-reorder and bulk delete: fields can be reordered by dragging and deleted in bulk. Field order drives how columns and form inputs are presented, so adjusting it no longer means deleting and recreating fields.
- Impact shown before deletion: deleting a table or field lists the configurations it affects first, so bindings do not silently fail afterwards.
Database
- Grant roles directly: roles can be added to an account from account data management, so testing permissions no longer requires writing an actionflow to grant them.
- Vacuum full: physically rebuilds a table to fully reclaim storage, which is worth running after large deletions. The table is locked while it runs, so schedule it during low-traffic hours.
Custom type and enums
- Custom types added: declare and use your own structured types. Define one once and reuse it across APIs, the data model, actionflows and AI agents. Read the docs
- Enum types added: declare and use enum types. Read the docs
Resources
- Resource management: browse images, videos and files with size and reference information, down to the referring table and field, to judge whether a file is safe to delete. Read the docs
- One-click cleanup of unreferenced resources: delete everything unreferenced in one pass to reclaim storage.
Action
Actionflow
- Nodes:
- Four new nodes: delay, throw error, while loop and break. Read the docs
- Input and output nodes support arrays and object types.
- Data query nodes now require a limit: without one, only aggregate fields are available, which prevents accidental full-table scans from overloading a live database.
- Debugging: configure inputs and run a whole actionflow, with database writes rolled back afterwards, so you do not have to keep cleaning up test data.
- Runtime rebuilt: the execution path was rewritten, with more accurate error messages and no silently swallowed exceptions.
- External access: every actionflow gets its own GraphQL API, so outside systems and your own front end can trigger it directly. Read the docs
API
External API integration was rewritten as a whole: importing, testing and configuring responses no longer means entering everything by hand. Read the docs
-
Import: import from OpenAPI documents (
.yaml/.json) or a cURL command, bringing in a whole set of APIs at once instead of entering them one by one. When an import fails, AI can analyze the cause. -
Requests and responses:
- Expanded to 5 request methods: GET, POST, PUT, DELETE and PATCH.
- Expanded to 3 body formats:
application/x-www-form-urlencodedandmultipart/form-datajoinapplication/json, so form submissions and file uploads can be integrated. - Responses branch by status code: configure a separate response structure per status code, with
2XXwildcards and a default fallback. APIs that return different shapes on success and failure no longer need manual branching and parsing in an actionflow. - Testing inside the editor: send a request and see status code, duration and content type; merge the result into a chosen response status or overwrite it, or generate a structure straight from raw JSON instead of typing fields.
-
Organization and usage:
- APIs live in collections, and each collection has its own variables (base URL, token and so on) shared by the whole set. APIs can move between collections.
- Pagination: turning it on generates page-index and page-size system inputs. Bind them to the API’s pagination parameters and a front-end list pages automatically with “Load more”.
- Responses bind directly: an API response can be bound straight to components, or passed to an actionflow or to AI.
Verification code, sign in and payment
- Custom verification code delivery: integrate your own SMS or email channel such as Twilio, bind delivery to your own actionflow, and customize content, signature, template and variables. This gets you past the platform’s SMS restrictions, message wording stays on brand, and several delivery methods can be configured at once for better delivery rates.
- Email verification code sign-in: end users can sign in with an email address and a verification code, without a phone number.
- SSO user-info authentication method: SSO gained a setting for how user info is authenticated, for identity providers with specific requirements at that step. Read the docs
- Airwallex payments added: accept payments through Airwallex, covering more currencies and regions. Read the docs
Developer access
A Momen backend is a self-documenting GraphQL service. This release opens it up properly: the top bar gains Developer access, where you pick up the endpoint and credentials needed to program against your project’s backend without going through the editor — or hand that work to an AI coding tool. Read the docs
- Endpoint and access token: the GraphQL endpoint, the subscription endpoint and an Admin Bearer Token. External clients send the token with each request. The API is self-documenting, and its shape tracks changes to your project automatically.
- GraphiQL playground: debug GraphQL queries inside the editor, with no separate client to install.
- Edit a project from a coding agent: integrate through the Momen plugin or MCP, covering Claude Code, Cursor (including team marketplaces), OpenCode and other MCP-capable tools.
Other improvements
- User center and project detail rebuilt: project list, project detail and account settings have been reorganized.
- Cloning and template creation no longer block the front end: creation runs in the background instead of holding you on a waiting screen.
- New
regexReplaceformula: regular-expression replacement, so string cleanup and formatting no longer need a run-code node. Read the docs
Bug fixes
Over 200 bugs fixed in this release.