PaperBanana
Open source implementation and extension of Google Research’s PaperBanana for automated academic figures, diagrams, and research visuals, expanded to new domains like slide generation.
What is PaperBanana?
PaperBanana is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to open source implementation and extension of google research’s paperbanana for automated academic figures, diagrams, and research visuals, expanded to new domains like slide generation.
Open source implementation and extension of Google Research’s PaperBanana for automated academic figures, diagrams, and research visuals, expanded to new domains like slide generation.
This server falls under the Data Science & ML category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- Open source implementation and extension of Google Research’
Use Cases
Maintainer
Works with
Installation
PIP
pip install paperbananaManual Installation
pip install paperbananaConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use PaperBanana
PaperBanana is an open-source multi-agent system for generating publication-quality academic figures, methodology diagrams, and research visualizations from text descriptions using large language and vision models. It extends Google Research's original PaperBanana concept with iterative VLM critic feedback loops, batch processing, venue-specific style packs (NeurIPS, ICML, ACL, IEEE), and an MCP server interface that lets AI assistants in Claude Code or Cursor generate and refine figures directly in chat. Researchers use it to automate the figure creation pipeline — from raw method descriptions to polished, ready-to-submit diagrams — without manually operating design tools.
Prerequisites
- Python 3.10+ with pip installed
- An OpenAI API key (OPENAI_API_KEY) or a Google AI Studio key (GOOGLE_API_KEY) for Gemini — at least one is required
- uvx (from the uv package manager) for the recommended MCP installation, or pip for CLI use
- An MCP-compatible client such as Claude Desktop or Cursor
Install PaperBanana via pip
Install the base package. For MCP usage, install with the mcp extra. For full provider support including Google Gemini, add the google extra.
pip install "paperbanana[mcp,google]"Set your LLM provider API key
Export the API key for your chosen provider. OpenAI is the default; Google Gemini (gemini-2.5-flash) is a free alternative.
# OpenAI (default)
export OPENAI_API_KEY="sk-..."
# OR Google Gemini (free tier available)
export GOOGLE_API_KEY="your-google-ai-studio-key"Add the MCP server to your client configuration
Configure your MCP client to launch the paperbanana-mcp server using uvx. This approach avoids manual virtual environment management.
{
"mcpServers": {
"paperbanana": {
"command": "uvx",
"args": ["--from", "paperbanana[mcp]", "paperbanana-mcp"],
"env": {
"OPENAI_API_KEY": "sk-..."
}
}
}
}Restart your MCP client
Restart Claude Desktop or reload the editor extension so it discovers the new paperbanana server. The MCP server exposes tools like generate_diagram, generate_plot, evaluate_diagram, batch_diagrams, and orchestrate_figures.
Test with a diagram generation request
Ask your AI assistant to generate a methodology diagram from a text description. The multi-agent pipeline (Retriever, Planner, Stylist, Visualizer, Critic) will iterate until the output meets quality criteria.
PaperBanana Examples
Client configuration
MCP config block for Claude Desktop using uvx to run the PaperBanana MCP server with an OpenAI key.
{
"mcpServers": {
"paperbanana": {
"command": "uvx",
"args": ["--from", "paperbanana[mcp]", "paperbanana-mcp"],
"env": {
"OPENAI_API_KEY": "sk-your-openai-key-here"
}
}
}
}Prompts to try
These prompts show how to generate methodology diagrams, plots, and evaluate figures using the MCP tools.
- "Generate a methodology diagram for a Retrieval-Augmented Generation pipeline with the caption 'Our RAG Framework Overview' in NeurIPS style"
- "Create a bar chart comparing model accuracy from my results.csv file with intent: accuracy comparison across baselines"
- "Polish this existing diagram figure.png for submission to ICML — make the arrows thicker and increase font size"
- "Evaluate this generated diagram against the reference human-drawn figure and give me a quality score"
- "Batch generate all figures described in my manifest.yaml file"Troubleshooting PaperBanana
uvx command not found when launching the MCP server
Install uv first: curl -LsSf https://astral.sh/uv/install.sh | sh. Then verify with uvx --version. Alternatively, install paperbanana with pip and use the command 'paperbanana-mcp' directly.
API key authentication errors from OpenAI or Google
Verify that OPENAI_API_KEY or GOOGLE_API_KEY is set in the env block of your MCP config (not just as a shell export, since MCP clients launch servers in isolated environments). Check that your key has active credits.
Diagram generation times out or returns a blank image
The multi-agent pipeline (Planner + Critic loop) can take 60-120 seconds for complex figures. Increase your MCP client's tool timeout if possible. For faster results, use --auto flag equivalents which skip manual critique cycles, or switch to the Google Gemini provider which has lower latency on flash models.
Frequently Asked Questions about PaperBanana
What is PaperBanana?
PaperBanana is a Model Context Protocol (MCP) server that open source implementation and extension of google research’s paperbanana for automated academic figures, diagrams, and research visuals, expanded to new domains like slide generation. It connects AI assistants to external tools and data sources through a standardized interface.
How do I install PaperBanana?
Install via pip with: pip install paperbanana. Then configure your AI client to connect to this MCP server.
Which AI clients work with PaperBanana?
PaperBanana works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is PaperBanana free to use?
Yes, PaperBanana is open source and available under the MIT license. You can use it freely in both personal and commercial projects.
PaperBanana Alternatives — Similar Data Science & ML Servers
Looking for alternatives to PaperBanana? Here are other popular data science & ml servers you can use with Claude, Cursor, and VS Code.
Ultrarag
★ 5.6kA Low-Code MCP Framework for Building Complex and Innovative RAG Pipelines
RocketRide
★ 3.1k📇 🏠 - MCP server that exposes RocketRide AI pipelines as t
Aix Db
★ 2.1kAix-DB 基于 LangChain/LangGraph 框架，结合 MCP Skills 多智能体协作架构，实现自然语言到数据洞察的端到端转换。
NeMo Data Designer
★ 1.9k🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data.
MiniMax
★ 1.5kBridges MiniMax AI capabilities to the Model Context Protocol, enabling AI agents to perform image understanding, text-to-image generation, and speech synthesis. It provides a standardized interface for accessing MiniMax's core tools via JSON-RPC.
NpcPy
★ 1.4kThe python library for research and development in NLP, multimodal LLMs, Agents, ML, Knowledge Graphs, and more.
Browse More Data Science & ML MCP Servers
Explore all data science & ml servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up PaperBanana in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use PaperBanana?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.