GraphXR Agent GraphXR Agent is a built-in AI assistant for working with graph data in a GraphXR project. Use it to turn natural-language questions and instructions into guided graph exploration, data preparation, and analysis. The Agent helps analysts move from a question to an actionable graph workflow by suggesting the next operation, explaining graph patterns, and helping prepare data for detailed analysis. Treat generated suggestions as analytical guidance and verify results against your source data. The Agent is available by default to any individual GraphXR user. Users with administrator privileges can install specific LLM models, and manage usage levels and per-user budgets. GraphXR Agent is currently in beta release. Features and behavior may change as the product evolves. GraphXR Agent Quick Start To get started using GraphXR Agent: Open a GraphXR project. Open GraphXR Agent from the left menu bar and click the New thread button. Input the prompt Create a random graph and click Send. Wait for the result. Your result may vary. By default, the Agent saves a view before any destructive code action. For examples of Agent workflows for exploring sample data, try a GraphXR Sandbox demo. Using GraphXR Agent You can use GraphXR Agent as soon as you log in and open a project. Its system prompt includes default canvas context, project context, and interaction mode. A gxr.agent API surfaces sessions, snapshots, and view-copying. You can: Set the agent to work on complex exploratory tasks using natural language prompts. Copy Agent code, and Agent responses as .md files to the clipboard. Review skills and tools available to the Agent. Manage chat threads. Manage views created by the Agent. Review current LLM usage and chat options. Review current usage budgets (in terms of daily token and dollar amounts) Show or hide the Agent per project using the Project > Settings > UI Configuration interface. Open The Agent Log in to GraphXR Create or open a project. Click the Agent icon on the left menu bar to open the Agent panel. You can right-click and choose Float to un-dock any panel from the left menu and reposition it conveniently on the project canvas. Create a New Thread A thread is the record of a chat session, including your prompts and the Agent responses (code and generated text explanations). Threads persist in the project until you delete them. A thread is labeled with the text of the first prompt you enter (e.g. list my skills), but it will likely include many other prompts. For example, in a single thread you could enter prompts to create graph data, query a connected database, find paths, lay out the graph in a specific pattern, create and save a grove notebook, and more. You can create a new thread at any time. To create a thread: Click the New Thread + icon at the top left. A New Thread tab is created. Enter a prompt (for example, list my tools) and click the Send icon. The thread is now labeled with the initial prompt text. Rename a Thread A thread is labeled with the text of its first prompt (e.g. please lay out Order nodes in a circle). You can rename it to be shorter, or more descriptive of the workflow in the thread. To rename a thread: Click the Settings icon at the top right, and select Rename Enter the new name in the dialog, and click Save. Close or Open an Existing Thread Open threads are shown in separate tabs in the Agent panel. The Thread History menu lists all the threads you’ve created (whether currently open or not) in order of when each was last opened. To close a thread: Click the X icon on the thread tab. The thread is closed (but not deleted). To open a closed thread: Click the Thread History icon at the upper right. Optionally, use the search bar to search for a thread by name. Select the thread from the menu. The thread is opened in a new tab. Delete Threads You can delete one thread, or clear all threads in a single step. To delete one thread: Open the thread you want to delete. Click the Settings icon at the top right. Select Delete thread To clear all threads: Click the Settings icon at the top right. Select Clear threads Click OK (or Cancel) in the dialog to confirm (or cancel) the action. Attach Files to a Thread You can attach files with data or information (text or image) that you want the Agent to evaluate in the context of the current prompt. To include an external file in a prompt: Click the Attach File icon at the bottom left of the panel. Navigate to the file and click Open to add it to the prompt. Enter the prompt (e.g. Find UI elements in the attached image) and click Send. The Agent evaluates the attached file according to the prompt and returns its response. Copy Agent Results Each code block or text response includes an icon to copy results to the clipboard to transfer threads to other GraphXR projects, or for use in reports. To copy Agent code or text responses: Click the copy icon next to a code block or a response. A code block is copied as code, and a response as an .md file. To copy an entire thread to the clipboard: Click the Settings icon at the top right. Choose Copy Thread from the menu. Start or Stop Automatic View Saving By default, the Agent saves a view before any destructive code action. If you find that too many views are being created, you can turn this function off. To stop automatic saving of views: Click the Settings icon at the top right. Choose Settings from the menu. In Save Checkpoint before agent runs code, click the button to turn off saving. Click the return arrow at the top left to exit Settings. Review LLM Usage and Budget Status As a user with administrator access (either individual, or in a multi-user installation), you can include LLMs to be used, and limit daily and overall budgets in terms of both tokens and dollar amounts. Setting budgets is done in the Admin’s Agent panel. Within a project, you can review current LLM and chat options, view usage status and add a model at any time. To review installed LLMs and chat options: At the bottom of the panel, the LLM menu shows the currently installed LLMs and highlights the one currently in use (e.g. Gemini Flash 3 Preview). Click the Chat Options icon next to the LLM menu to see current thinking level and debug options. At any time, you can: Select a different thinking level from the menu. Click the Debug logging radio button to turn logging on (or off). To review LLM usage: Click the Settings icon at the top right. Choose Settings from the menu. The Usage section shows Tokens used today and Dollars used today across all projects for the logged in user. Click the return arrow at the top left to exit Settings. To add an LLM model: Click the Settings icon at the top right. Choose Settings from the menu. Click Add model and enter the required Basic and Advanced model details. Click Submit to add the model (or Cancel to exit). Click the return arrow at the top left to exit Settings.