ParaView-MCP: An Autonomous Visualization Agent with Direct Tool Use

TL;DR

ParaView-MCP addresses the high barrier to using advanced visualization tools by coupling a multimodal language model with ParaView through the Model Context Protocol, enabling natural-language interaction and viewport-grounded decision making. The system combines a curated set of ParaView functions exposed via an MCP server with a dedicated ParaView Manager and a viewport-aware feedback loop, forming a hybrid API-vision agent. It demonstrates workflows such as language-driven parameter tuning, visualization by example, and cross-tool collaboration, supported by preliminary domain expert feedback. The work highlights a practical, extensible path toward autonomous visualization agents that can democratize access to SciVis tools and enable collaborative analytics across tools, with potential extensions to batch processing and plug-in ecosystems.

Abstract

While powerful and well-established, tools like ParaView present a steep learning curve that discourages many potential users. This work introduces ParaView-MCP, an autonomous agent that integrates modern multimodal large language models (MLLMs) with ParaView to not only lower the barrier to entry but also augment ParaView with intelligent decision support. By leveraging the state-of-the-art reasoning, command execution, and vision capabilities of MLLMs, ParaView-MCP enables users to interact with ParaView through natural language and visual inputs. Specifically, our system adopted the Model Context Protocol (MCP) - a standardized interface for model-application communication - that facilitates direct interaction between MLLMs with ParaView's Python API to allow seamless information exchange between the user, the language model, and the visualization tool itself. Furthermore, by implementing a visual feedback mechanism that allows the agent to observe the viewport, we unlock a range of new capabilities, including recreating visualizations from examples, closed-loop visualization parameter updates based on user-defined goals, and even cross-application collaboration involving multiple tools. Broadly, we believe such an agent-driven visualization paradigm can profoundly change the way we interact with visualization tools. We expect a significant uptake in the development of such visualization tools, in both visualization research and industry.

Figures (4)

• Figure 1: The agent automatically identifies the iso-value that leads to half of the original surface area.
• Figure 2: Iterative transfer function design driven by a vision-based feedback loop. The agent starts with the default colormap (Cold-to-Warm) to assess the value distribution, and based on user instruction (green tree, brown base), designs and refines the colormap by visually examining each rendering output.
• Figure 3: An illustration of the co-exploration use case, where the ParaView-MCP works with a novice user by suggesting a list of possible visualization methods, explaining specific techniques (e.g., streamline), and guiding the overall exploration process.
• Figure 4: A multi-tool collaboration example use case, where Paraview-MCP and Blender-MCP ahujasid_blender_mcp work together to create an illustration from a volumetric dataset of a tooth. Per user prompt, ParaView-MCP extracts the iso contour and then renders the surface in Blender-MCP based on the example image the user provided.