惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

C
Check Point Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
L
LangChain Blog
云风的 BLOG
云风的 BLOG
M
MIT News - Artificial intelligence
A
About on SuperTechFans
J
Java Code Geeks
量子位
博客园 - 三生石上(FineUI控件)
博客园 - Franky
博客园_首页
H
Hackread – Cybersecurity News, Data Breaches, AI and More
IT之家
IT之家
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
Apple Machine Learning Research
Apple Machine Learning Research
Engineering at Meta
Engineering at Meta
雷峰网
雷峰网
D
DataBreaches.Net
人人都是产品经理
人人都是产品经理
Martin Fowler
Martin Fowler
有赞技术团队
有赞技术团队
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
I Audited 3 AI Coding Tools for Privacy — The Difference ...
yurenpai · 2026-06-18 · via DEV Community

yurenpai

Claude Code stores 43MB of your conversations. Cursor saves your passwords in plaintext JSONL. Cline is the only one that puts your API key in the OS keychain — but it still archives everything you say.

Not one of them tells you this is happening.


Why I Did This

Two weeks ago, I discovered Claude Code had silently archived 43MB of my complete conversation history on my machine. I wrote about it. The comments asked: "What about Cursor? What about Cline?"

I tested them. The results were more nuanced than I expected.


Methodology

Three tools. Five dimensions. One standardized test script.

Dimension What I Tested
Conversation Storage What's stored? Format? Size?
Credential Security How is the API key stored? Plaintext or encrypted?
User Awareness Does the tool tell you it's archiving your conversations?
Data Isolation Can AI in Project A read conversations from Project B?
Uninstall Residue Does conversation data survive uninstallation?

In each tool, I typed the same three things: a normal coding question, a database connection string with a password, and an API key.


Results

Claude Code — The Hoarder

Conversation Storage: ~/.claude/projects/, JSONL format. One file per session.

Opening one 2.3MB session file:

Content Count
AI responses 590
AI internal thinking 272
Tool calls 101
Tool call results (incl. file paths) 100
File history snapshots 208

Credential Storage: settings.json — plaintext. The built-in Read tool can access this file without restriction.

Data Isolation: No. Global history file merges all projects.

Uninstall Residue: 19 project directories, 43MB of conversation data.


Cursor — Stores More Than You Think

Conversation Storage: agent-transcripts/, JSONL format. Same structure as Claude Code — user input, AI responses, tool calls all stored. Size depends on conversation length.

Stores every user input, every AI response, and every tool it called. Does not store tool execution results or thinking blocks — in Plan mode.

Credential Storage: SQLite database stores tokens at 415 bytes each. Better than plaintext.

My test password and API key: Both written verbatim into the transcript file.

Data Isolation: Yes. Per-project separation.


Cline — The Only One That Got Credentials Right

Conversation Storage: Open source, directly auditable. Each task gets its own directory with 4 JSON files:

tasks/<taskId>/
  api_conversation_history.json   — Full API messages (incl. tool results)
  ui_messages.json                — Chat UI records
  task_metadata.json              — Metadata
  context_history.json            — Context evolution

Stores no less than Claude Code in conversation content — complete API message format, tool use and tool results included.

Credential Storage: The only tool that got this right. API key goes through VS Code Secrets API to Windows Credential Manager. Not stored in a file.

Data Isolation: Yes. Per-task isolation.


Comparison Table

Dimension Claude Code Cursor Cline
User input Full text Full text Full text
AI responses Full text Full text Full text
AI internal thinking 272 entries None None
Tool execution results Includes file paths Plan mode: no Yes
Session size 2.3MB 16KB 4 JSON files
API Key storage Plaintext JSON SQLite OS keychain
Conversation format JSONL JSONL JSON
User informed No Asked about sharing No
Cross-project isolation No Yes Yes
Open source (auditable) No No Yes

Ranking

Credential security:  Cline >>> Cursor >> Claude Code
Conversation privacy: All three archive everything, none tells you
Data volume:          Claude Code >>>>> Cline > Cursor
Transparency:         Cline (open source) > Cursor > Claude Code

No tool scored perfectly.


What This Really Reveals

Three tools. Three architectures. Two business models. One thing they all share:

They all archive your conversations in local plaintext. Not one tells you.

This isn't one tool's vulnerability. This is the AI coding tool category defaulting to "archive everything" — and users not knowing it.


What You Can Do

Level Action
Daily Periodically clean up conversation directories
Never Paste real passwords or tokens into AI conversations
Before selling Delete all tools' conversation archives
When choosing Know what each tool stores before you install

Part 3 of the "AI Tool Privacy" series. Next: Prompt Injection in Practice — How a Malicious CLAUDE.md Can Make Your AI Betray You.