# NVEIL [Health: Active]

**Category:** 📊 Data Visualization  
**Repository:** https://github.com/nveil-ai/nveil-toolkit  
**GitHub Stars:** 3  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/nveil

## Description
Data processing and visualization toolkit — 50+ chart types, raw data stays local.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "nveil": {
    "command": "uvx",
    "args": ["nveil"],
    "env": {
      "NVEIL_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `NVEIL_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What NVEIL MCP server does

The NVEIL MCP server exposes NVEIL's data processing and visualization workflow to MCP clients. An agent can describe a desired chart or analysis in natural language, using datasets available to the local environment. NVEIL plans the required operations and visualization, then the toolkit executes those operations locally.

The workflow covers more than basic chart generation. Supported use cases include joins, aggregations, pivots, time-series processing, geospatial heatmaps, 3D volumes, scientific plots, medical imaging, biosignal data, network graphs, and other visualization types. NVEIL supports more than 50 visualization types and can select Plotly for 2D output, VTK for 3D or medical data, and DeckGL for geospatial visualizations.

## How it works

The NVEIL MCP server uses a two-stage process. First, NVEIL AI receives dataset metadata rather than the raw records. The documented metadata includes column names, data types, and aggregate statistics. NVEIL AI returns a processing and visualization plan based on the user's request.

Next, the local toolkit carries out the plan. Joins, transformations, aggregations, pivots, and rendering happen on the user's machine. The result can be displayed as a figure or exported in supported formats. This separation is useful when raw data must remain local while an agent still needs help selecting a visualization and planning data operations.

Generated specifications are deterministic and can be saved as `.nveil` files. A saved specification can later be loaded and rendered against fresh data without another API call, which supports repeatable reports and offline re-rendering.

## Setup and configuration

Install the Python package with:

```bash
pip install nveil
```

Python 3.10 or newer is required. Create an account at `app.nveil.com`, generate an API key in Settings, and provide it through the `NVEIL_API_KEY` environment variable or the Python configuration API.

To launch the MCP interface, run:

```bash
nveil mcp
```

The command starts a stdio server for an MCP client. A client configuration can invoke the `nveil` executable with `mcp` as its argument. The README specifically documents use with Claude Code, Claude Desktop, Cursor, and Codex.

## Tools and capabilities

The available workflow supports:

- Natural-language requests for charts and data processing
- Local execution of joins, aggregations, pivots, and rendering
- Automatic selection among Plotly, VTK, and DeckGL backends
- Publication-ready HTML and PNG output through the toolkit's command-line workflow
- Reusable `.nveil` specifications that render on new datasets
- Visualization of 2D, 3D, geospatial, scientific, medical, biosignal, and network data

The MCP interface is a fit for agents that need to inspect or transform local datasets and produce visual output without loading raw records into the remote planning service.

## Limitations and notes

An NVEIL API key and account are required for the planning stage. The privacy model keeps raw data local, but metadata is sent to NVEIL AI. The README identifies this metadata as column names, types, and aggregate statistics, so deployments should evaluate whether those details are acceptable for their data.

Rendering a generated specification is local and requires no API call. However, creating a new specification uses the NVEIL planning service. The MCP server is launched over stdio, so the selected MCP client must be able to start a local command and provide access to the datasets used by the request.

NVEIL is licensed under GNU AGPL v3 or later. The project also states that commercial dual licensing is available through the vendor.

_Full upstream README: https://allmcps.com/mcp/nveil/readme_