PaperBanana

PaperBanana is an agentic framework and MCP server that generates publication-quality academic diagrams and statistical plots from text descriptions using a multi-agent pipeline with GPT or Gemini vision models, including iterative auto-refinement.

Evaluated Mar 06, 2026 (0d ago) vlatest

Homepage ↗ Repo ↗ Other mcp academic diagrams visualization research figures openai gemini multiagent

⚙ Agent Friendliness

/ 100

Can an agent use this?

🔒 Security

/ 100

Is it safe for agents?

⚡ Reliability

/ 100

Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality

Documentation

Error Messages

Auth Simplicity

Rate Limits

🔒 Security

TLS Enforcement

Auth Strength

Scope Granularity

Dep. Hygiene

Secret Handling

Document/paper management MCP. Local document access. PDF/document contents may be confidential. No remote auth for local use.

⚡ Reliability

Uptime/SLA

Version Stability

Breaking Changes

Error Recovery

Best When

A researcher needs to produce polished methodology or results figures for a paper draft without manual design tools, and wants AI-driven iterative refinement.

Avoid When

Cost of LLM API calls per diagram is prohibitive, or highly customized branding/style requirements exist.

Use Cases

• Generating methodology flowcharts and architecture diagrams for AI research papers from plain-text descriptions
• Creating statistical plots and charts from CSV/JSON experimental results data
• Auto-refining generated figures iteratively until a VLM critic deems them publication-ready
• Integrating diagram generation directly into Claude Code or Cursor workflows via MCP

Not For

• Non-academic or marketing design work
• Interactive or animated data visualization
• Teams without OpenAI or Gemini API access and budget

Interface

REST API

GraphQL

gRPC

MCP Server

Yes

SDK

Yes

Webhooks

Authentication

Methods: api_key

OAuth: No Scopes: No

Requires OpenAI API key, Google Gemini API key, or Azure OpenAI credentials via .env file or CLI flags.

Pricing

Model: open_source

Free tier: No

Requires CC: Yes

Tool itself is open-source; underlying model API costs apply. Gemini free tier provides a cost-free option.

Agent Metadata

Pagination

none

Idempotent

Retry Guidance

Not documented

Known Gotchas

⚠ Unofficial community implementation not affiliated with original paper authors
⚠ Requires Python 3.10+ and specific model API keys configured before use
⚠ Auto-refine mode can consume many API calls in a single run
⚠ Three MCP tools: generate_diagram, generate_plot, evaluate_diagram - agents should call evaluate after generate

Alternatives

diagrams-net matplotlib plotly inkscape napkin-ai

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for PaperBanana.

$99

API endpoint ↗ Agent guide ↗ Report inaccuracy

Scores are editorial opinions as of 2026-03-06.