# Version 2.4.2 | Oct 14, 2025

## **Overview** <a href="#overview" id="overview"></a>

ZBrain Builder 2.4.2 release brings enhanced flexibility, productivity, and user experience to the platform. Key updates include per-agent model configurations in the Agent Crew, import/export of Agent Crews, multi-document management in the Agent Dashboard, and integration of the ZBrain MCP server. Flows gain an expanded, Markdown-enabled editor for clearer system instructions, while the new Nano Banana image generation model expands creative options with enterprise-grade performance. Together, these updates make ZBrain Builder more adaptable and impactful for real-world AI workflows.

#### ZBrain Builder 2.4.2 release overview <a href="#zbrain-builder-2.4.2-release-overview" id="zbrain-builder-2.4.2-release-overview"></a>

| Component      | Capability                                             | What it delivers                                                                                                                                          |
| -------------- | ------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Flows & Pieces | Expanded system instructions & prompt modal for models | Delivers a larger, editable view for system instructions, allowing easier review and minimal scrolling                                                    |
|                | Markdown support in system instructions                | Delivers clear, editable, and well-structured instructions through preserved formatting and an integrated Markdown editor                                 |
| Integrations   | Nano Banana model integration                          | Delivers a new image generation option with high accuracy, enterprise-grade performance, and compatibility with ZBrain’s full image generation capability |
| RAG            | Customizable automated reasoning variables & rules     | Delivers precise, context-aware outputs by enabling users to define and tailor variables and rules before LLM processing                                  |
| Agents         | Multi-document management in the agent dashboard       | Enables uploading multiple documents that delivers efficient document comparison and context switching, improving analysis and productivity               |
|                | Agent crew import & export                             | Delivers portable, reusable, and restorable agent crew configurations through JSON-based import/export                                                    |
|                | Per-agent model configuration in the agent crew        | Delivers precise control over model selection, enabling tailored agent performance                                                                        |
|                | ZBrain MCP integration in agent crews                  | Delivers streamlined crew configuration through reusable, pre-configured MCPs with full compatibility                                                     |

## New features <a href="#new-features" id="new-features"></a>

### Flows & Pieces <a href="#flows-and-pieces" id="flows-and-pieces"></a>

#### Markdown support for system instructions & prompt box <a href="#markdown-support-for-system-instructions-and-prompt-box" id="markdown-support-for-system-instructions-and-prompt-box"></a>

ZBrain Builder now supports proper Markdown formatting in the system instructions & prompt box within Ask AI. Content output from LLMs preserves its formatting automatically, and users can apply or edit formatting either by typing Markdown syntax or by using an integrated Markdown editor with rich text options.

**Navigation:** Flows → Ask AI Model→ Model Configuration → System Instructions

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2FjNtrFiWcY1hXUmniWYJB%2Fimage.png?alt=media&#x26;token=75c4848f-112e-4bad-a418-14e0c77f8bdc" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Auto-preserves formatting from pasted LLM-generated instructions.
* Supports live Markdown rendering while typing syntax (e.g., `**Bold**`, `# Header`).
* Enables structured, readable, and well-formatted system instructions and prompts without manual cleanup.

### Integrations <a href="#integrations" id="integrations"></a>

#### Nano Banana model for image generation <a href="#nano-banana-model-for-image-generation" id="nano-banana-model-for-image-generation"></a>

ZBrain Builder now includes the **Nano Banana** model (Gemini 2.5 -flash) as part of its image generation models. Users can select this model directly from the interface or via API to generate high-quality, visually accurate images, with performance aligned to enterprise SLAs.

**Navigation:** Flows → Ask AI Model → Model Selection → gemini 2.5-flash

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2FSDKSyXZRMycZDgQqF9Vx%2Fimage.png?alt=media&#x26;token=ba377823-1bea-4492-a9b9-6260f3b343e2" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Expands creative flexibility with an additional model tailored for diverse image generation needs.
* Improves efficiency with high-quality image outputs delivered within SLA response times.
* Strengthens enterprise readiness with compatibility across UI, API, and batch workflows.
* Supports scalability by integrating seamlessly with existing image generation models.

### **Agents** <a href="#agents" id="agents"></a>

#### Import & export of agent crew configurations <a href="#import-and-export-of-agent-crew-configurations" id="import-and-export-of-agent-crew-configurations"></a>

ZBrain Builder now allows users to export agent crew configurations as JSON files and import them back when needed. This enables seamless backup, sharing, and restoration of crews across projects and environments.

**Navigation:**

* **Export:** Agents → Define Crew Structure → Settings Icon → Export → Save the file
* **Import:** Agents → Define Crew Structure → Settings Icon → Import → Select the file

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2FlSypJiWOAirQTlWWDtbK%2Fimage.png?alt=media&#x26;token=6f487d10-34e1-4c29-b1f5-9bd875fe3259" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Enhances portability and reusability by allowing crews to be easily moved across workspaces or projects.
* Improves collaboration by allowing teams to share standardized crew setups.
* Strengthens resilience with quick restoration of crews from backups.
* Increases productivity by eliminating the need to recreate complex crew structures manually.

#### Per-agent model configuration in agent crews <a href="#per-agent-model-configuration-in-agent-crews" id="per-agent-model-configuration-in-agent-crews"></a>

ZBrain Builder now supports per-agent model configuration within agent crews, allowing each agent to leverage the most suitable LLM for its role. This provides greater flexibility, performance optimization, and cost control compared to a single crew-wide model configuration.

**Navigation:** Agents → Define Crew Structure → Select agent->Agent Side Panel → Model

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2FDHhFJ55sEwoj7KoqLXhp%2Fimage.png?alt=media&#x26;token=2803c1d8-9fdb-43b1-a9b6-bf062061a99b" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Enhances flexibility by assigning different models to different agents within the same crew.
* Optimizes performance and cost by matching model choice to agent roles (e.g., reasoning-heavy vs. utility agents).
* For older crews and agents, the Crew-Level model config acts as the default fallback for agents without explicit configurations.

#### ZBrain MCP integration in agent crews <a href="#zbrain-mcp-integration-in-agent-crews" id="zbrain-mcp-integration-in-agent-crews"></a>

ZBrain Builder now allows users to configure agent crews with existing ZBrain MCPs, making it easier to leverage pre-configured capabilities without duplication. Users can select from available ZBrain MCPs, assign them to agents, and seamlessly use them alongside external MCPs within the same crew.

**Navigation:** Agents → Define Crew Structure → Agent Side Panel → Add MCP Server → ZBrain MCP Servers

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2FEOcZoWP6K0LORsCbi7kG%2Fimage.png?alt=media&#x26;token=044b425a-84d6-47cf-a387-92c71e6e1a40" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Increases efficiency by reusing pre-configured ZBrain MCPs.
* Enhances flexibility by combining ZBrain MCPs with external MCPs in the same crew.
* Improves usability by attaching clear tool descriptions to each MCP for better context.
* Reduces setup time and complexity when designing advanced multi-agent workflows.

### **RAG** <a href="#rag" id="rag"></a>

#### Customizable knowledge variables & rules in automated reasoning <a href="#customizable-knowledge-variables-and-rules-in-automated-reasoning" id="customizable-knowledge-variables-and-rules-in-automated-reasoning"></a>

ZBrain Builder now allows users to add or edit variables and rules sent to the LLM through a unified **Reasoning Prompt Box**. This gives users greater flexibility to tailor reasoning outputs to specific business contexts and requirements.

**Navigation:** Login → Knowledge → Automated Reasoning

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2FRed0IFTxU0bup1Ke1NuX%2Fimage.png?alt=media&#x26;token=7f965347-d5c3-41bb-b3ba-6ee7a2e12b9a" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Improves accuracy by letting users provide contextual details (e.g., distinguishing between similarly named files).
* Enhances flexibility with customizable prompts for both Knowledge Variables and Extracted Rules.
* Saves time by generating Knowledge Variables and Extracted Rules in parallel once both prompts are provided.
* Optimizes control with default model selection and per-task model overrides.

## **Improvements** <a href="#improvements" id="improvements"></a>

### Flows & Pieces <a href="#flows-and-pieces.1" id="flows-and-pieces.1"></a>

#### Expanded system instructions & prompt modal <a href="#expanded-system-instructions-and-prompt-modal" id="expanded-system-instructions-and-prompt-modal"></a>

ZBrain Builder now supports an expanded modal view for the system instructions & prompt box in the Ask AI Model configuration in Flows. Users can click the expand arrow to open a larger, scrollable modal that displays full context, making it easier to review and edit instructions without constant scrolling.

**Navigation:** Flows → Ask AI → Model Configuration → System Instructions → Click Expand

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2Fxr8FaWIeaKv7yHZyTPIL%2Fimage.png?alt=media&#x26;token=20b4764d-44cd-474a-9919-970a500604a1" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Views most or all instructions in a single, enlarged modal to minimize scrolling.
* Edits instructions directly within the expanded modal, with auto-save enabled.
* Preserves text formatting (Markdown, line breaks, spacing) for improved clarity.
* Scrolls vertically when content exceeds the screen size.
* Returns to the compact view seamlessly by closing the modal.
* Enlarges the modal without triggering the data selector, ensuring a smooth editing experience.

### Agents <a href="#agents.1" id="agents.1"></a>

#### Multiple document uploads in the agent dashboard <a href="#multiple-document-uploads-in-the-agent-dashboard" id="multiple-document-uploads-in-the-agent-dashboard"></a>

ZBrain Builder now supports uploading multiple documents in the **Agent Dashboard**, enabling users to seamlessly analyze, compare, and switch between up to 10 documents within a single session.

**Navigation:** Agents → Create New / Select Existing Agent → Agent Dashboard → Upload

<figure><img src="https://3781630280-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIbuSicczDKTyHzwzicar%2Fuploads%2FyVprV18xockiYOkJ7k2P%2Fimage.png?alt=media&#x26;token=0af1f663-b36b-437b-9940-2bda7f3e6999" alt=""><figcaption></figcaption></figure>

**Key outcomes:**

* Saves time by eliminating the need for repeated uploads when working with multiple documents.
* Improves productivity by allowing seamless comparison and analysis within a single interface.
* Enhances decision-making with clearer context switching across related documents.
* Strengthens accuracy by ensuring activity logs and outputs remain tied to the correct document.