ParaView-MCP (LLNL)
MCP serverCommunity
A research MCP server from LLNL that lets a multimodal model drive ParaView and check its own results by looking at the viewport. By Lawrence Livermore National Laboratory.
71 starsUpdated 7 months agoGitHub data from 5 Nov 2026
At a glance
- Type
- MCP server
- Works with
- ParaView
- How it connects
- Drives a running copy of the tool
- Access
- Can modify models
- Tool versions
- unknown (ParaView from conda-forge)
- Operating systems
- Not stated in its documentation
- Needs a ParaView licence
- No
- Transport
- Local (stdio)
- AI clients documented
- Claude Desktop
- Licence
- BSD-3-Clause
- Language
- Python
What it can do
- Creates and adjusts visualisations from natural language and image input.
- Observes the viewport and refines the visualisation iteratively.
- Accompanies an IEEE VIS 2025 short paper describing the approach.
Limitations
- The authors warn that it relies on pvserver and client synchronisation, which is deprecated in recent ParaView versions, so the GUI may not show content correctly.
- You start pvserver with multiple clients and connect the GUI by hand before using it.
Install
git clone https://github.com/LLNL/paraview_mcp.git
cd paraview_mcp
conda create -n paraview_mcp python=3.10
conda install conda-forge::paraview
conda install mcp[cli] httpxMCP servers and skills run code on your computer. Review the source before installing.
AI-generated models, drawings and results must be checked by a qualified engineer before use.