PaperBanana
MCP server and CLI that turns paper text or data into academic figures, diagrams and plots
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Runs code and calls an external generation API
- Consumes paid Google quota
Install
Manual install
{
"mcpServers": {
"paperbanana": {
"command": "uvx",
"args": ["--from", "paperbanana[mcp]", "paperbanana-mcp"],
"env": { "GOOGLE_API_KEY": "your-google-api-key" }
}
}
}Runs the server via uvx with no local clone.
This is third-party code. Review the repository files before installing.
What it does
PaperBanana assembles figures for research work: methodology diagrams from a text description, statistical plots from CSV or JSON, and full figure packages for a whole paper. It plugs into an agent over MCP via uvx, and also runs as a standalone command and as a set of Claude Code skills. Over MCP it exposes generation and refinement of diagrams and plots, evaluation against a reference, batch runs and reference downloads. Rendering runs on Google Gemini models, so a Google key is required.
Who it is for. For researchers, analysts and paper authors who need diagrams and plots for publications and talks.
Good fit when
- You need a methodology diagram from a text description
- You need a statistical plot from a data table
- You need a consistent set of figures for a paper
Not a fit when
- You have no access to Google Gemini or a key
- You need UI or layout design rather than scientific figures
Example request
Build a methodology diagram from this experiment description and a plot from results.csvLimitations
Generation requires a Google Gemini key, and Google access is blocked from Russia without a VPN. Some features need optional extras (studio, pdf), and the benchmark reference set weighs a few hundred megabytes.
How to disable. Remove the paperbanana entry from your MCP config, or uninstall the package: pip uninstall paperbanana.
MCP
- Transport
- stdio
- Authentication
- API key
| Environment variables | |
|---|---|
| GOOGLE_API_KEY required, secret | Google key for the Gemini models used to render images |
Security check
- Runs code and calls an external generation API
- Consumes paid Google quota
README in short
The README presents an open-source implementation and extension of Google Research's PaperBanana for automated academic figures. Install via pip or from source, with Docker and a Colab notebook for a quick start. The MCP server connects over uvx and exposes eleven tools: generating and continuing diagrams and plots, evaluation, batch runs and reference downloads. The repo also ships three Claude Code skills and a Gradio studio mode. MIT licensed.
FAQ
Do I need a local clone?
No, the MCP server runs via uvx without cloning. For the CLI you can install the package via pip.
Which model renders?
Images are generated by Google Gemini models, so a Google key is needed.
Related
166 skills for scientific work: bioinformatics, cheminformatics, clinical data, geospatial analysis and 100+ databases
Google's open-source MCP server for databases: ready tools for Postgres, MySQL, BigQuery, Spanner and more, plus custom tools in tools.yaml
Official Hugging Face skills: Hub operations via the hf CLI, datasets, model training, Spaces, evals and deployment
A visualization language for agents and an MCP server: neat charts from a simple semantic spec