For the Reusable Generator approach, the Agent now prompts you to confirm that you want to use AI-generated text for the values. After the script is written, regular value generation does not incur token cost. AI-generated text does.
Fixed an issue with memory errors for longer generations.
Reusable Generator data generation approach - As of September 4, 2026, the Reusable Generator approach is out of private preview and publicly released. The Reusable Generator approach automatically creates a workflow to complete the requested data generation. This approach is especially useful to generate extremely large volumes of data. With the workflow, you can use the workflow to quickly create a smaller set of data to verify the overall structure before you generate the full set of data. The Data Agent can automatically switch to the Reusable Generator approach based on your request. You can also switch to the Reusable Generator approach from the mode selection dropdown list in the Data Agent prompt. The dropdown now includes both modes (Plan, Agent, Autonomous) and approaches (Single Dataset and Reusable Generator). Single Dataset is the default approach, which generates a single-use script to complete the task.
Workspace notifications - Within a workspace, you can now configure a list of notification destinations. A notification destination can be an email address, a Slack incoming webhook URL, or a Microsoft Teams incoming webhook URL. When a workflow job completes, fails, or is canceled, Fabricate sends a message to the enabled notification destinations for the workspace.
Project home page - The new project home page includes summary metrics for the project, the list of project databases and database connections, and the list of project workflows.
Account jobs - On the account home page, the navigation menu now contains a Jobs option to display the list of workflow jobs from across the account. The list includes the project name, workspace name, workflow name, and job status.
Using secrets from AWS Secrets Manager - A workspace secret can now be a secret from AWS Secrets Manager.
Default account models - Account owners can now specify the default models to use for data generation, validation, and freeform text generation. The new settings are on the AI page of Account Settings.
API key AI models and improved edit panel - The API key configuration now includes the models to use for data generation, validation, and freeform text generation when using the API key for calls to the MCP server. By default, API keys use the default models for the account. A new consolidated edit panel makes it easier to change the configuration of an API key.
Improved caching to reduce the number of tokens used.
You can now delete attachments from the Data Agent chat.
Account roles for API keys - You can now assign an account role (Owner, Auditor, Member) to an account API key, to manage the available actions that you can perform using the key. The API creation panel includes the role selection, and you can also change the assigned role for existing API keys.
The new Auditor account role allows an account user or API key to view all account workspaces, account settings, audit log events, and usage settings.
Improved the email message template that displays on mobile devices.
Improved support for PDF fonts.
Fixed an issue with creating .zip files for download.
External profiling skill - Added a new database profiling skill that you can install into an AI agent such as Claude or Cursor. You use this skill to create a profile of a database that is not connected to Fabricate. You can then add the profile to Fabricate to use for data generation. For example, you might want to create synthetic data that uses the same structure as a production database, but you do not want to connect the database to Fabricate.
You can now select GPT-5.6 Luna as the model to use for freeform text generation.
Made some small display improvements to the user interface.
Improved handling and display of domain names. Includes a new Allowed Domains tab on the account details panel displayed from the Admin Console.
When you configure your account to synchronize SSO groups, you can also choose to automatically create a new workspace for each new SSO group. The workspace has the same name as the group. You also select the workspace role to assign to the group.
A reference for the Fabricate OpenAPI API is now available at https://fabricate.tonic.ai/api.
The MCP server now provides access to account owner options to track account usage, AI usage costs, and audit logs.
Fixed an issue where an error was returned if an Agent conversation did not end with a user message.
Fixed an issue where OAuth failed when the users's preferred model was retired.
You can now use assumed roles for connections to AWS. On the workspace settings page, you set up a list of available IAM roles for the workspace. When you configure a server connection to PostgreSQL or MySQL, you can choose to use a role for authentication. The role trust policy must allow the Fabricate's principal to call sts:AssumeRole. On a self-hosted instance, you also set the environment variable FABRICATE_AWS_ASSUMER_ROLE_ARN to the assumer role for your Fabricate instance.
You can now use Azure service principals for connections to Azure-hosted databases. On the workspace settings page, you set up a list of available service principals for the workspace. Your Azure database must contain a Microsoft Entra ID user that is mapped to this service principal and that is granted the required permissions. When you configure a server connection to PostgreSQL, MySQL, or SQL Server, you can choose to use a service principal for authentication.
When you export a local database, you can now choose whether to export the entire database or only the currently selected table.
On the free tier, users who sign up with a business email address now receive a larger token allowance than users who sign up with personal email address.
Amazon Redshift support - You can now create server connections to Amazon Redshift.
Consistent UI for in-process server connection creation - Made the UI consistent for creating new server connections during database export and database connection creation. For both processes, you can now choose whether to add the new connection to the workspace server connection list.
Using CA certificates for connection authentication - You can now set up a list of available CA certificates for your account. On the server connection details, the TLS setting is replaced with an SSL/TLS dropdown. For PostgreSQL, Amazon Redshift, and MySQL, you can choose to use CA verification.
Fixed an issue with validating the form for pushing a database export to a database.
Fixed an issue where the Agent forgot uploaded files after it completed a context summarization.
The Data Agent now supports Claude Sonnet 5.
You can now display the timestamp for chat messages. To display the timestamp, hover the mouse over the message while you press the Shift key.
Fabricate now queues chat messages, so that you can continue to type new messages as the Agent processes previous ones.
Fixed an issue to ensure that the Agent has the correct database size limits.
Improved support for very large exports to Google BigQuery.
Other fixes and internal enhancements.
Validation agent - Plus and Enterprise users can now enable the validation agent for a Data Agent chat. The validation agent can analyze the quality of the Data Agent scripts (whether they will generate quality data) and the generated data itself. By default, validation is off. From the chat prompt, you can enable validation and set whether to check the scripts or only the data. Note that enabling validation increases the time to completion and the cost of the Data Agent requests.
Fixed an issue where the Data Agent provided incorrect guidance.
Fixed an issue where summarization of Data Agent conversations failed.
Other internal bug fixes and enhancements.
Local database size limit - Fabricate now enforces a limit on the volume of data in an individual local database, based on the license plan.
- Free plan - 256 MB
- Plus plan - 1 GB
- Enterprise plan - 5 GB
Fabricate displays a warning when the database size reaches 80% of the allowed maximum. Existing databases that exceed the limit are not truncated, but you can no longer add data to them.
Support for Azure Blob Storage - You can now set up Fabricate to allow the Data Agent to read from and write to Azure Blob Storage. You set up your Azure account as an allowed domain for the account, and the credentials (either an account key or a SAS token) as a workspace secret.
Added missing descriptions for audit log events.
Fixed an issue with workflows.
Fixed an issue where reverting to a previous point in one conversation deleted attached files in other conversations.
Project cloning - You can now create a clone of a project. When you clone a project, you can select the workspace for the new project, and choose whether to copy the chat history.
Added effort level to model selection - When you select the model for the Data Agent to use, you can also select the level of effort.
The Data Agent can now automatically spin up a sub-agent to perform and manage tasks. It currently uses a sub-agent for profiling.
Fixed a scrolling issue with the login page display.
Fixed an issue where exports to PDF failed because of unsupported fonts.
Fixed an issue where Fabricate attempted to use read-only mode on databases that did not support it.
Fix an issue where mock API request headers were not forwarded.