Methodology
The Key Safety Score
A 0 to 100 rating of how well an app protects the API key you give it. Rules-based, published and applied to every listing.
Scoring rubric
| Factor | Max | Scoring |
|---|---|---|
| Key storage | 40 | Local (on-device, direct to provider) 40. Vendor server 18. Undocumented 10. |
| Auditable code | 25 | Open source 25. Closed source 0. |
| Self-hosting | 15 | Self-hostable service 15. Local desktop, CLI or editor app 8. Vendor-hosted web app 0. |
| Offline option | 10 | Supports local models (Ollama, LM Studio) 10. |
| Provider choice | 10 | 2 points per supported provider, up to 10. |
Grades: A is 80 or above, B is 60 to 79, C is 40 to 59, D is below 40. The score covers key handling only. It is not a security audit and says nothing about output quality. Inputs come from public documentation, source code where available, and our own testing. Makers can submit corrections.
A
190 tools
B
95 tools
C
6 tools
D
7 tools
Highest scoring tools
- 1LobeChat (LobeHub) · Chat Clients100
- 2SillyTavern · Creative & Roleplay100
- 3AnythingLLM · Notes & Knowledge100
- 4big-AGI · Chat Clients100
- 5CrewAI · Automation & Agents100
- 6GPT Researcher · Research & Search100
- 7marimo · Coding Agents & IDEs100
- 8NextChat · Chat Clients100
- 9PDFMathTranslate · Translation100
- 10bolt.diy · Coding Agents & IDEs100
- 11Browser Use · Automation & Agents100
- 12Jupyter AI · Coding Agents & IDEs100
- 13Next AI Draw.io · Creative & Roleplay100
- 14OpenClaw · Automation & Agents100
- 15Plandex · Coding Agents & IDEs100
- 16ONLYOFFICE AI Agent · Writing100
- 17Project AIRI · Creative & Roleplay100
- 18Refact.ai · Coding Agents & IDEs100
- 19VideoLingo · Translation100
- 20Word GPT Plus · Writing100