paperbanana

Overview

PaperBanana is an agentic framework for generating publication-quality academic diagrams and statistical plots from text descriptions. It implements a multi-agent pipeline with iterative refinement, supporting multiple VLM and image generation providers including OpenAI, Azure OpenAI, and Google Gemini. The server offers features like auto-refine mode, batch generation from manifest files, and integration with PDF inputs for methodology context.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• generate

  Generate methodology diagrams from text descriptions with optional optimization and refinement

  Try:Generate a diagram for our encoder-decoder architecture with sparse routingTry:Create a methodology diagram for the transformer model with auto-refine modeTry:Continue generating with feedback to make arrows thicker and colors more distinct
• plot

  Generate statistical plots from data files with communicative intent

  Try:Create a bar chart comparing model accuracy across benchmarks using results.csvTry:Generate a scatter plot showing the relationship between training epochs and accuracy
• batch

  Generate multiple diagrams from a manifest file with optional composite figures

  Try:Generate multiple diagrams from batch_manifest.yaml with optimizationTry:Create batch diagrams with composite layout for the panels
• plot-batch

  Generate multiple statistical plots from a manifest file

  Try:Generate multiple plots from plot_batch_manifest.jsonTry:Create batch statistical plots for different benchmark results
• batch-report

  Generate human-readable reports from existing batch runs

  Try:Generate a markdown report from batch run outputs/batch_20250109_123456_abcTry:Create an HTML report from a specific batch ID
• sweep-report

  Generate reports from sweep experiments showing ranked variants

  Try:Generate an HTML report from sweep outputs/sweep_20250109_123456_abcTry:Create a markdown report from a specific sweep ID

Note: Tool names inferred from CLI commands, though not explicitly listed as MCP tools. PaperBanana does support MCP server functionality but the README doesn't clearly document the MCP API schema.

Comparable tools

Installation

pip install paperbanana

{
  "mcpServers": {
    "paperbanana": {
      "command": "python",
      "args": ["-m", "paperbanana.mcp"],
      "env": {
        "OPENAI_API_KEY": "your-api-key"
      }
    }
  }
}

FAQ

Which LLM providers are supported?

PaperBanana supports OpenAI (GPT-5.2 + GPT-Image-1.5), Azure OpenAI / Foundry, and Google Gemini providers.

Can I generate diagrams from PDF papers?

Yes, PaperBanana supports PDF inputs with optional per-page selection for methodology context.

On Hacker News

Recent discussion from the developer community.

• Show HN: Implementation of Google's PaperBanana (diagram generation from text)▲ 3💬 1
  Story by dippatel1994 · 2026-02-04
• Show HN: Open-source PaperBanana – academic diagrams from text via agents▲ 1
  Story by dippatel1994 · 2026-02-06

Compare paperbanana with