--- title: "9.2.0 Overview" slug: "9-2-0-overview" status: "update" updated: 2026-11-05T21:47:16Z published: 2026-11-05T21:47:16Z canonical: "docs.unqork.io/9-2-0-overview" --- > ## Documentation Index > Fetch the complete documentation index at: https://docs.unqork.io/llms.txt > Use this file to discover all available pages before exploring further. # 9.2.0 Overview > [!WARNING] > The 9.2.0 release schedule differs from previous releases. Staging and UAT deploy on the same day, and Production follows one week later. | Environment | Release Date | | --- | --- | | Staging | Nov 12, 2026 | | UAT | Nov 12, 2026 | | Production | Nov 19, 2026 | Version 9.2.0 adds Build Agent controls, supports MCP in two directions, scopes private Smart Components to a module, and improves Module Builder and Explorer. ## About This Overview This page summarizes 9.2.0 before the full release notes are available. Use it to explore what is coming and to plan for the release. It lists highlights only. The full 9.2.0 release notes publish on Nov 5, one week before Staging. They cover every improvement and fix, plus the Key Updates that can affect existing applications. ## UnqorkAI > [!WARNING] > These highlights are available to customers with an active AI capabilities agreement. Contact an Unqork representative to learn about activating AI-powered features. ### Build Agent **Control who uses the Build Agent.** Administrators select **Administration** > **Environment Settings** > **Build Agent** > **Access Control**. They set each role to **No Access**, **Ask Only**, or **Builder**. With **Ask Only**, the Build Agent answers questions about a workspace but cannot change it. Access can differ by environment and workspace. **Limit each run.** The **Limits** page caps **LLM calls per run** and **Run time limit (minutes)**, with defaults of 500 calls and 60 minutes. When a run reaches a limit, the Build Agent pauses, and Creators select **Continue** to resume. Changes to these settings are recorded in audit logs. **Shared context across tasks.** The Build Agent now keeps documents for a workspace, starting with an architecture document for each application. A plan for a new application begins with a task that writes that document, and later tasks read it before planning. Creators read, edit, and delete these documents on the **Context** tab next to **History**. The Build Agent checks content before saving it and rejects credentials and other sensitive values. A gauge in the prompt box displays how full the agent's working space is. **Better building blocks.** The Build Agent now reads formulas, exact values, and numeric formats from attached Excel workbooks. It can then implement the calculations in a rater or pricing model. It builds login, registration, and group management flows for Express end-users. ### Working with MCP The Model Context Protocol (MCP) is an open standard that lets AI agents use external tools and data. Version 9.2.0 supports MCP in two directions. **Build Agent to external systems (MCP Connectors).** Administrators connect the Build Agent to systems outside Unqork on the **MCP Connectors** page under **Build Agent**. They choose a connector from the catalog: Atlassian, Google Drive, Google Docs, Exa, or Firecrawl. They then select which of its tools the Build Agent can use. For example, an administrator can connect Google Docs so the Build Agent works from a requirements document while it plans an application. Each Creator signs in to the connector with their own account from **Manage Connectors** in the Build Agent composer menu. Creators then set each connection to **Ask**, which pauses for approval on every call, or **Trusted**, which runs without asking. **External AI agents to Unqork (MCP Server, Beta).** Published API modules become tools that external AI agents call at one endpoint, `/api/1.0/mcp`. In an API module's settings, an administrator turns on **Agents can call this module as a tool (MCP)** on the **MCP Settings** page. For example, a company assistant outside Unqork can call a published quote-calculation API module as a tool instead of rebuilding that logic. An agent signs in with an API credential. Its Express role needs Write on the module's **User Permissions**, so existing permissions still decide what the agent can reach. ### Smart Components **Private Smart Components are scoped to a module.** A public Smart Component is scoped to the environment, and a private one is scoped to a module. Before 9.2.0, a module could place a private component only once. Now a module can place it any number of times. The component is versioned with the application, promoted with the module, copied when Creators duplicate or paste the module, and rolled back with the module. Names must be unique only in the module. **Move from private to public.** **Make Public** creates a public Smart Component from a private one. Creators name the public component and choose which instances in the module switch to it. Public Smart Components stay environment-wide and reusable across modules. Restoring a module revision brings its private components back and does not roll back public ones. **Simpler editing.** Private Smart Components no longer have drafts. Changes the Build Agent makes to a private component save straight into its module. The **Components** tab lists a module's private components in an **In This Module** group. The Smart Components Hub groups them by workspace, application, and module. ### Chat Agent Smart Component **One place for conversation logging.** Before 9.2.0, each Chat Agent Smart Component had its own conversation logging settings. Now the **Conversation audit module** field on the LLM Config sets the module for every Chat Agent Smart Component that uses that LLM Config. Each conversation is one submission. Components that relied on the old settings stop recording until Creators set the audit module. A blank field records no conversations. **Visualizations from data questions.** Creators turn on **Enable Generated Visualizations**, which is off by default. The Chat Agent Smart Component can then answer a data question with a visualization that it writes at runtime. The platform validates the visualization on the server first. The visualization displays in that conversation only and is not saved. The data comes from the API modules approved in **Server Modules**. An empty list means the component reads only the live form. **Server-side runtime.** The Chat Agent Smart Component now runs on the server with its LLM Credential, and end-user behavior stays the same. Its **LLM Config** and **LLM Credential** must be fixed values. ## Unqork Platform These highlights are available to all customers. **Module Builder.** **Save** now stays unavailable until Creators apply or discard all draft settings changes. A tooltip names the components with unapplied changes, because saving first could lose configuration. With **Debug** on, the Unit Tests panel for a Data Workflow component displays a step-by-step Debug Trace. Each operator's input and output has an expandable JSON view. Module Builder also uses less processing time with large modules. **Explorer.** The new **Config Search** tab searches component settings across an application. Creators pick an application, and a branch or version for versioned applications, then enter at least 4 characters. Search finds whole-word, case-insensitive matches and returns up to 1,000 components. **API Debugger.** The **Docs** tab has a **Copy link** button that copies a link to the selected documentation. Opening the link displays that endpoint in the API Debugger, and the link survives sign-in. **Platform fixes.** Version 9.2.0 resolves issues across Application Versioning, Workflow, Workspaces, and Module Builder. ## Before Upgrading Some changes in 9.2.0 can affect an existing application, a permission, or a setting. Administrators and QA teams can plan for them now. Examples include: - Chat Agent Smart Components stop recording conversations until Creators set the audit module. - A module can place the same private Smart Component more than once, and restoring a module revision restores its private components. - The Navigation component **Hide Previous**, **Hide Next**, **Hide Submit**, and **Hide Draft** settings now apply in Express. The full release notes list every change of this kind when they publish on Nov 5. --- **Changelog** | Date | Change | | --- | --- | | 2026-11-05 | Initial publication. | The Staging environment is the first level of application building in Unqork. Use it to build, test, and style your applications. Use the UAT (user acceptance testing) environment to realistically test and navigate your applications as end-users would use them. The Production environment is the final stage of the development process. Use this environment to promote tested and functioning application so your end-user can interact with it. Large Language Models (LLM) are artificial intelligence (AI) programs that use deep learning algorithms to analyze and generate text and other content based on large amounts of data. A definition Also known as Unqork Users, or Designer Users; is anyone who is inside the Unqork platform. The Unqork front-end application framework that delivers Creator-built applications to end-users. Express is the layer end-users interact with—the published forms, screens, and navigation built in the Module Builder. Express is distinct from Express View, which is a Creator-facing preview tool in the IDE. End-users, also known as Express Users, are the individuals accessing an application through Express View. In most cases, end-users are the customers using the product. Also known as Unqork Users, or Designer Users; is anyone who is inside the Unqork platform. APIs (application programming interfaces) are a set of protocols and definitions developers use to build and integrate application software. APIs act as the connective tissue between products and services. An API endpoint is the location where an API receives requests for a specific resource on its server. It's the point of contact between an API client and an API server, and is usually a uniform resource locator (URL) that provides the location of the resource. End-users, also known as Express Users, are the individuals accessing an application through Express View. In most cases, end-users are the customers using the product. JSON (JavaScript Object Notation) is an open standard file and data interchange format. Unqork uses JSON for submission (record) data. In Unqork an application is a program used to perform specific tasks. An application can be one of two types: workflow or module. Use your QA (quality assurance) environment to test applications to ensure reliability, security, and functionality before releasing them to the end-user.