NVEIL MCP server

<!-- mcp-name: io.github.nveil-ai/nveil -->

NVEIL Toolkit

Describe your data. Get production charts. Your data stays local.

Quickstart &bull; API Reference &bull; Examples &bull; Changelog

NVEIL is an AI-powered data visualization toolkit. Write one line of natural language, and NVEIL processes your data and generates publication-ready visualizations — no chart code, no hallucinations, no data leaving your machine.

import nveil

nveil.configure(api_key="nveil_...")

# Pass a file path directly — no DataFrame loading required.
spec = nveil.generate_spec("Revenue by region, colored by quarter", "sales.csv")

fig = spec.render("sales.csv")   # 100% local — no API call
nveil.show(fig)                   # opens in browser

From your shell

After pip install nveil the nveil command is on your $PATH:

export NVEIL_API_KEY=nveil_...

# Ground yourself on the dataset (shape / dtypes / head preview)
nveil describe sales.csv

# Generate HTML + PNG + a reusable .nveil spec, print the explanation
nveil generate "Revenue by region, colored by quarter" \
  --data sales.csv --format all --explain

# Re-render an existing spec on fresh data — no API call
nveil render chart.nveil --data new_sales.csv

For AI agents (Claude Code / Claude Desktop / Cursor / Codex / …)

NVEIL ships first-class integrations:

# Claude Code / Claude Desktop — install the bundled skill
nveil install-skill

# Claude Desktop, Cursor, any MCP client — add an MCP server:
# {"mcpServers": {"nveil": {"command": "nveil", "args": ["mcp"]}}}
nveil mcp                    # stdio server; launched by the MCP client

Why NVEIL?

¹ ChatGPT Advanced Data Analysis, Claude Analysis tool, Gemini Data Agent &nbsp;·&nbsp; ² PandasAI, LIDA, Julius, Vanna &nbsp;·&nbsp; ³ Plotly, Matplotlib, Seaborn

How It Works

Your Data ──> Toolkit ──metadata only──> NVEIL AI ──> Processing Plan ──> Local Execution ──> Result
               ^                                                           ^
          raw data stays here                                     raw data stays here

Shortened. The full README is on GitHub.