Perfect memory for Claude - Never solve the same coding problem twice
Project description
Claude Cache ๐ง
claude
โโโโโโโ โโโโโโ โโโโโโโโโโ โโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโ โโโโโโโโโโโ
โโโ โโโโโโโโโโโ โโโโโโโโโโโโโโ
โโโ โโโโโโโโโโโ โโโโโโโโโโโโโโ
โโโโโโโโโโโ โโโโโโโโโโโโโโ โโโโโโโโโโโ
โโโโโโโโโโ โโโ โโโโโโโโโโ โโโโโโโโโโโ
Your personal coding intelligence that learns from both successes AND failures. Claude Cache captures your complete problem-solving journey - what works, what doesn't, and the path between them.
๐ง Learning from Everything, Storing What Matters
Claude Cache revolutionizes how AI coding assistants remember. Unlike traditional tools that only save working code, we capture your complete learning journey:
โ Success Patterns - What Works
- Elegant solutions that worked first time
- Proven approaches you can trust
- Best practices from your actual experience
๐ซ Anti-Patterns - What Doesn't Work
- Failed approaches to avoid
- Dead ends that waste time
- Context-specific pitfalls you've discovered
๐บ๏ธ Journey Patterns - The Path to Solution
- Complete problem-solving sequences
- Key insights that unlock breakthroughs
- The "why" behind the solution
Why This Changes Everything
Consider this real scenario:
Monday: You spend 2 hours debugging an authentication loop
- Try localStorage fix (fails)
- Check cookies (fails)
- Debug state management (fails)
- Finally realize it's a useEffect issue (success!)
Friday: You hit the same auth loop issue
Without Claude Cache: You might try localStorage again, wasting time on approaches that didn't work.
With Claude Cache:
- โ ๏ธ Immediately warns: "Don't check localStorage - that failed before"
- โ Suggests: "This is likely a useEffect issue - here's what worked"
- ๐บ๏ธ Shows the journey: "Last time this took 4 attempts, skip to solution #4"
๐ฏ Pattern Classification System
Every pattern is classified by quality and type:
| Type | Symbol | Description | Example |
|---|---|---|---|
| Gold | ๐ | Worked first time, elegant | "useEffect with cleanup solved it immediately" |
| Silver | ๐ฅ | Worked after 2-3 attempts | "Third approach with useMemo worked" |
| Bronze | ๐ฅ | Eventually worked | "Finally solved after trying 5 approaches" |
| Anti-Pattern | ๐ซ | Confirmed failure | "localStorage doesn't work for auth tokens" |
| Journey | ๐บ๏ธ | Complete problemโsolution path | "Auth fix: tried A, B failed โ C worked because..." |
| Caution | โ ๏ธ | Works but has tradeoffs | "Quick fix but needs refactoring" |
Native Claude Code Tools
Type / in Claude Code to access these powerful tools:
/mcp__cache__query
Search your entire knowledge base instantly
Example: /mcp__cache__query "authentication JWT"
Returns: Your previous JWT implementations with context
/mcp__cache__learn
Save successful solutions for future use
Example: /mcp__cache__learn
solution: "Fixed CORS with proxy middleware"
tags: "cors,api,middleware"
/mcp__cache__suggest
Get proactive recommendations based on current context
Example: /mcp__cache__suggest "working on API endpoints"
Returns: Relevant patterns from your knowledge base
/mcp__cache__stats
Monitor your growing knowledge base
Shows: Total patterns, projects, search capabilities
/mcp__cache__browse
Index documentation for instant access
Example: /mcp__cache__browse "https://docs.example.com"
Result: Documentation indexed and searchable
๐ Quick Start (2 Minutes)
1. Install Claude Cache
# Using pipx (recommended - isolated installation)
pipx install "claude-cache[mcp]"
# Alternative: Using pip directly
pip install --user "claude-cache[mcp]"
# Basic version
pipx install claude-cache
2. First Run - Choose Your Mode
Real-time Learning (Recommended):
cache start --watch
One-time Processing:
cache start
You'll see:
โ Knowledge base initialized at ~/.claude/knowledge/cache.db
Processing 47 existing log files...
Processing: YourProject/session1.jsonl
Processing: YourApp/session2.jsonl
...
โ Learned 23 patterns from your history
Starting real-time monitoring...
What's happening?
- Claude Cache found all your past Claude Code sessions
- It's analyzing your historical coding patterns
- Learning from your successes AND failures
- Building your personal knowledge base
3. Use It Immediately
# Interactive monitoring dashboard (press 'h' for help, 't' for tutorial)
cache monitor
# Query your knowledge
cache query "authentication"
# Get contextual suggestions
cache suggest
# See what you've learned
cache stats
3. Claude Code Integration (Optional)
Add to your .claude.json:
{
"mcpServers": {
"cache": {
"type": "stdio",
"command": "cache-mcp"
}
}
}
Start MCP server separately:
cache-mcp
Restart Claude Code and type / to see your new tools!
4. Test It's Working
# Check status
cache stats
# Search existing patterns
cache query "authentication"
# Get suggestions
cache suggest --context "working on APIs"
๐ฌ How It Works
Claude Cache uses dual-path learning to build comprehensive intelligence:
1. Historical Analysis (First Run)
- Processes all your existing Claude Code sessions
- Extracts patterns from past successes and failures
- Builds initial knowledge base from your history
- Takes 30 seconds to 10 minutes depending on history size
2. Real-Time Monitoring (Continuous)
- Watches your coding sessions in real-time
- Captures successful solutions ("perfect!", "works!")
- Learns from failures (errors, "broken", "failed")
- Records complete problemโsolution journeys
- Zero lag - runs silently in background
3. Intelligent Retrieval
- Semantic Search: ML-powered understanding of meaning
- Pattern Matching: Finds similar problems across projects
- Context Awareness: Right solution for your specific stack
- Journey Replay: Shows the path that worked before
4. Privacy First Design
- 100% Local: All data stays on your machine
- No Cloud: Never uploads or shares your patterns
- No Telemetry: We don't track what you're learning
- You Own It: SQLite database you fully control
๐ What Gets Captured?
โ Automatically Captured:
- Solutions that work (you say "thanks", "perfect", "works")
- Approaches that fail (errors occur, you say "broken", "failed")
- Complete problemโsolution journeys
- Patterns that repeat across projects
- High-confidence solutions (80%+ success rate)
- Consistent failures (repeatedly don't work)
โ Not Captured:
- Every single edit or keystroke
- Low-confidence patterns
- Project-specific hacks
- Trial-and-error noise
- One-off solutions
Perfect For
- Solo Developers: Build a personal knowledge base of solutions
- Development Teams: Share successful patterns and best practices
- Learning: Capture and revisit complex problem-solving approaches
- Productivity: Eliminate repetitive problem-solving across projects
๐ก Real-World Impact
After using Claude Cache for a month, developers typically have:
- 50-100 high-quality patterns per project
- 30-50 anti-patterns preventing repeated mistakes
- 20-30 journey patterns showing problem-solving paths
- 90%+ relevance rate when patterns are suggested
- 50% reduction in time spent on familiar problems
Example Output
$ cache query "auth redirect"
๐ Found 3 relevant patterns:
1. โ
Success Pattern (React)
"useEffect cleanup prevents auth redirect loops"
Project: YourApp | Confidence: 92%
2. ๐ซ Anti-Pattern
"Don't store auth tokens in localStorage"
Failed in: YourApp, AnotherApp | Alternative: Use httpOnly cookies
3. ๐บ๏ธ Journey Pattern
"Auth loop fix: localStorage โ โ cookies โ โ useEffect โ
"
Time saved next time: ~45 minutes
Performance
- Speed: <100ms query response for 10K+ patterns
- Accuracy: 60-90% relevance in semantic matching
- Storage: Efficient SQLite with optional vector embeddings
- Privacy: Zero external API calls, completely local
Terminal Usage
Claude Cache offers multiple ways to run in terminal:
๐ Quick Start (Recommended)
# Start background learning system
cache background
# Search patterns
cache query "authentication patterns"
# Get suggestions
cache suggest --context "working on APIs"
# View statistics
cache stats
โ๏ธ Advanced Options
# Full system with terminal interface
cache run
# Process existing logs only (one-time)
cache process
# Foreground mode (for testing)
cache start --watch
# Include MCP server
cache run --with-mcp
๐ Background Process Methods
# Using nohup (survives terminal closure)
nohup cache start --watch > cache.log 2>&1 &
# Using screen (detachable sessions)
screen -S claude-cache -d -m cache start --watch
# Using tmux (session management)
tmux new-session -d -s claude-cache 'cache start --watch'
๐พ Manual Learning
# Save successful solutions
cache learn "JWT middleware with validation" --tags "auth,jwt,security"
# Index documentation
cache browse https://docs.example.com
cache scan-docs . # Scan current repository
# Export/import knowledge
cache export backup.json
cache import backup.json
๐ ๏ธ Process Control
# Check what's running
ps aux | grep cache
# Stop background processes
pkill -f 'cache start'
# View logs
tail -f /tmp/claude-cache.log
๐ Complete guide: See docs/TERMINAL_SETUP.md for detailed setup options.
๐๏ธ Architecture
Claude Cache/
โโโ ๐ง Dual-Path Learning Engine
โ โโโ Success Pattern Detector
โ โโโ Anti-Pattern Analyzer
โ โโโ Journey Pattern Tracker
โ โโโ Pattern Classification System
โโโ ๐พ Knowledge Base (SQLite)
โ โโโ Success Patterns (Gold/Silver/Bronze)
โ โโโ Anti-Patterns (What Not to Do)
โ โโโ Journey Patterns (ProblemโSolution)
โ โโโ Cross-Project Intelligence
โ โโโ Documentation Index
โโโ ๐ Intelligent Search
โ โโโ Semantic Understanding (ML)
โ โโโ Context-Aware Matching
โ โโโ Pattern Similarity Analysis
โโโ ๐ MCP Integration
โ โโโ Native Claude Code Tools
โโโ ๐ Quality Control
โโโ Confidence Scoring
โโโ Pattern Validation
โโโ Continuous Evolution
๐ Documentation
- ๐ฆ Installation Guide - Step-by-step installation for all platforms
- ๐ Quick Start Guide - Get running in 2 minutes
- ๐ฃ๏ธ Communication - How Claude Cache provides feedback and guidance
- ๐ง Philosophy - Why dual-path learning matters
- ๐ง Configuration - Advanced options
- ๐ How It Works - Technical deep dive
- ๐ก Use Cases - Real-world examples
- ๐ฅ๏ธ Terminal Setup - Command line mastery
Contributing
We welcome contributions! Areas of interest:
- Additional pattern detection algorithms
- Better journey pattern analysis
- Support for more development environments
- Team collaboration features
- Language-specific pattern recognition
See CONTRIBUTING.md for guidelines.
License
MIT License - see LICENSE for details.
Acknowledgments
Built with love for the developer community. Special thanks to all early adopters and contributors who helped shape Claude Cache into what it is today.
๐ฏ The Ultimate Goal
Claude Cache aims to be your external coding brain - remembering not just what worked, but understanding:
- Why it worked
- When it works
- What doesn't work
- How you got there
Every developer's journey is unique. Claude Cache ensures that journey makes you smarter with every line of code you write.
Ready to never repeat the same mistake twice? Install Claude Cache today and build your personal coding intelligence.
"The best developers aren't those who never fail, but those who learn from every failure and success alike."
Claude Cache is an independent tool for enhancing Claude Code, not an official Anthropic product.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file claude_cache-0.8.2.tar.gz.
File metadata
- Download URL: claude_cache-0.8.2.tar.gz
- Upload date:
- Size: 184.1 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.10.18
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
c5d2aeeba09ea4231f63841f82d0e156cae32e688894ae2ebc6fb4f077adf9a0
|
|
| MD5 |
594c71b5699e3578be51213c35edd103
|
|
| BLAKE2b-256 |
777c174ffd586765c6f35d2cce402d992e94e322f3859da5dab99516b03d401d
|
File details
Details for the file claude_cache-0.8.2-py3-none-any.whl.
File metadata
- Download URL: claude_cache-0.8.2-py3-none-any.whl
- Upload date:
- Size: 132.2 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.10.18
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
c616aa7459644efc22130790e4105ecc7b3132fb75f5a14785eab544875b93c8
|
|
| MD5 |
0ffc9aaeccec1048881e80c2a17f8b81
|
|
| BLAKE2b-256 |
893368a20f4ce28ef92e4443968c1151a69f0d98e5dceb52bdaefb4117eb9a14
|