$ timeahead.in
/ servers/pypi/paperbanana
pypi

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.

2k stars2k/wkupdated 16d agogithub ↗
90good
▣ Overview

What it does

Generates publication-quality academic diagrams and statistical plots from text descriptions through a two-phase multi-agent pipeline. Supports OpenAI (GPT-5.2 and GPT-Image-1.5), Azure OpenAI, Google Gemini, and Atlas Cloud as image generation backends. Offers batch diagram generation from YAML/JSON manifests, batch plot generation from data descriptions, optional PDF context loading, and a local Gradio web UI called Studio. Integrates with Claude Code via MCP with skills for /generate-diagram, /generate-plot, and /evaluate-diagram.

Who it's for

Academic researchers and scientists authoring papers or presentations who need to generate high-quality figures programmatically. Suited for AI researchers working in Claude Code or other LLM-integrated environments, particularly those who lack design expertise or prefer code-driven illustration.

Common use cases

  • Generate methodology diagrams and architecture visualizations from text descriptions
  • Create statistical plots from CSV or JSON data specifications
  • Iteratively refine diagrams based on feedback via auto-refine mode
  • Batch-generate multiple diagrams and plots from a single manifest file
  • Include PDF methodology pages as context to improve diagram fidelity

Setup pitfalls

  • Requires Python 3.10 or higher
  • Requires an API key for image generation: OpenAI, Google Gemini (free tier available), Azure OpenAI, or Atlas Cloud
  • Classified high-risk due to filesystem read/write access and outbound network calls—ensure working directories are sandboxed when processing untrusted input
  • Keep API keys in environment variables, not configuration files
▣ Score BreakdownMCPScore = Σ(raw × weight)
DimensionRawWeighted
Security
35%
100
35.0
Freshness
25%
100
25.0
Adoption
20%
73
14.7
Quality
10%
100
10.0
Trust
10%
50
5.0
Total
89.7
⚿ Capabilities & Risk Explainer
fs readfs writenetworkexecsecrets
◆ Risk level: high· 5 tools · auth: API key
fs read + fs write + network + exec + secrets active — can execute code, access credentials, and make external network calls.
Tool nameDescriptionDestructive?
generate_diagramGenerate a methodology diagram from text context and caption✓ no
continue_runContinue refinement for an existing run directory (optional critic feedback)⚠ yes
generate_plotGenerate a statistical plot from JSON data and intent description✓ no
evaluate_diagramCompare a generated diagram against a human reference on 4 dimensions✓ no
evaluate_plotCompare a generated statistical plot against a human reference on 4 dimensions✓ no
⚙ Install config
Claude Desktop · Cursor · Windsurf · VS Code (Copilot) · Claude Code
add to your MCP client config:
{
  "mcpServers": {
    "paperbanana-1": {
      "command": "uvx",
      "args": [
        "paperbanana"
      ]
    }
  }
}
📈 Score historylast 45 snapshots
5/10/20267/6/2026 · 45 snapshots
⚙ Maintenance health
74/ 100 · is this project alive?
contributors (1y)35
top contributor share43%
releases (1y)6
last release24d ago
ci✓ passing
⛁ Raw data
weekly downloads2k
github stars2k
forks302
open issues7
license✓ present
readme length35811 chars
last publish0d ago
last commit16d ago
last updated20h ago
install verified✓ pass · 49d ago
owner of this server? claim your listing to get a verified badgeclaim →
🔔 Score drop alerts
get notified by email when this server's score drops 5+ points