Context that persists for Claude Code - Knowledge Base, Task Management, and Conflict Detection
Project description
Clauxton
Context that persists for Claude Code
⚠️ Alpha Status: Clauxton is currently in Phase 0 development. Core features are being implemented. Not yet ready for production use.
Clauxton is a Claude Code plugin that provides persistent project context to solve AI-assisted development pain points.
Vision (Roadmap):
- ✅ Session Context Loss → Persistent Knowledge Base (Phase 0 - In Progress)
- 🔄 Manual Dependency Tracking → Auto-inferred task dependencies (Phase 1 - Planned)
- 🔄 Post-hoc Conflict Detection → Pre-merge conflict prediction (Phase 2 - Planned)
🎯 Quick Start
Note: CLI installation only. Full Claude Code plugin integration coming in Phase 1.
# Install from source (PyPI release coming soon)
git clone https://github.com/nakishiyaman/clauxton.git
cd clauxton
pip install -e .
# Initialize in your project
cd your-project
clauxton init
# Add knowledge to your Knowledge Base
clauxton kb add
# Search your Knowledge Base
clauxton kb search "architecture"
✨ Features
✅ Phase 0: Foundation (Complete)
Knowledge Base Management
- ✅ Persistent Context: Store architecture decisions, patterns, constraints, conventions
- ✅ Category System: Organize entries by type (architecture, constraint, decision, pattern, convention)
- ✅ YAML Storage: Human-readable, Git-friendly YAML format
- ✅ TF-IDF Search: Relevance-based search with automatic ranking (see Search Algorithm)
- ✅ CRUD Operations: Add, get, update, delete, list entries
- ✅ Atomic Writes: Safe file operations with automatic backups
- ✅ Secure Permissions: 700/600 permissions for privacy
🚧 Phase 1: Core Engine (In Progress - Week 6/8)
Knowledge Base CRUD (✅ Week 3 + Week 7 - Complete)
- ✅ MCP Tools: kb_search, kb_add, kb_list, kb_get, kb_update, kb_delete
- ✅ CLI Commands: kb add, kb get, kb list, kb search, kb update, kb delete
- ✅ Claude Code Integration: .claude-plugin/mcp-servers.json
- ✅ Type-Safe: Full Pydantic validation
- ✅ Version Management: Automatic versioning on updates
Task Management (✅ Week 4 - Complete)
- ✅ CRUD Operations: Add, get, update, delete, list tasks
- ✅ Dependency Tracking: Define task dependencies (DAG structure)
- ✅ Cycle Detection: Prevent circular dependencies with DFS algorithm
- ✅ Priority Management: Critical > High > Medium > Low
- ✅ AI Recommendations: get_next_task() based on priority and dependencies
- ✅ CLI Commands: task add, task list, task get, task update, task delete, task next
- ✅ YAML Persistence: tasks.yml with automatic backups
Task Management MCP Tools (✅ Week 5 - Complete)
- ✅ MCP Tools: task_add, task_list, task_get, task_update, task_next, task_delete
- ✅ Auto Dependency Inference: Infer dependencies from file overlap
- ✅ Claude Code Integration: Full task management via MCP
- ✅ AI-Powered Recommendations: get_next_task() via MCP
Dependency Analysis (⏳ Week 6 - Planned)
- 🔄 Task Graph Visualization: ASCII/Mermaid dependency graphs
- 🔄 Enhanced Inference: Multi-file pattern analysis
🔄 Phase 2: Conflict Prevention (Planned)
Pre-merge Conflict Detection
- 🔄 File Overlap Detection: Detect potential merge conflicts
- 🔄 Risk Scoring: Calculate conflict risk (0.0-1.0)
- 🔄 Safe Execution Order: Recommend optimal task order
- 🔄 Drift Detection: Detect scope expansion
📦 Installation
Requirements
- Python: 3.11 or higher
- Dependencies:
click>=8.0.0- CLI frameworkpydantic>=2.0.0- Data validationpyyaml>=6.0.0- YAML parsingmcp>=0.1.0- MCP server integrationscikit-learn>=1.3.0- TF-IDF search (optional, falls back to simple search if not installed)numpy>=1.24.0- Required by scikit-learn
Development Installation (Current)
# Clone repository
git clone https://github.com/nakishiyaman/clauxton.git
cd clauxton
# Create virtual environment
python3 -m venv venv
source venv/bin/activate # On macOS/Linux
# or
venv\Scripts\activate # On Windows
# Install in editable mode (includes all dependencies)
pip install -e .
# Verify installation
clauxton --version
PyPI Installation (Coming Soon)
# Full installation (with TF-IDF search)
pip install clauxton
# Minimal installation (simple keyword search only)
pip install clauxton --no-deps
pip install click pydantic pyyaml mcp
Note on Search: Clauxton uses TF-IDF algorithm for relevance-based search when scikit-learn is installed. If not available, it automatically falls back to simple keyword matching. See Search Algorithm for details.
🚀 Usage
Knowledge Base Commands (Phase 0 ✅)
# Initialize Clauxton in your project
clauxton init
# Add knowledge entry (interactive)
clauxton kb add
# Search Knowledge Base (TF-IDF relevance ranking)
clauxton kb search "architecture" # Results ranked by relevance
clauxton kb search "API" --category architecture
clauxton kb search "FastAPI" --limit 5 # Limit to top 5 results
# List all entries
clauxton kb list
clauxton kb list --category decision
# Get entry by ID
clauxton kb get KB-20251019-001
# Update entry
clauxton kb update KB-20251019-001 --title "New Title"
clauxton kb update KB-20251019-001 --content "New content" --category decision
# Delete entry
clauxton kb delete KB-20251019-001
clauxton kb delete KB-20251019-001 --yes # Skip confirmation
Task Management Commands (Phase 1 Week 4 ✅)
# Add a new task
clauxton task add --name "Setup database" --priority high
# Add task with dependencies
clauxton task add \
--name "Add API endpoint" \
--depends-on TASK-001 \
--files "src/api/users.py" \
--estimate 3.5
# List all tasks
clauxton task list
clauxton task list --status pending
clauxton task list --priority high
# Get task details
clauxton task get TASK-001
# Update task
clauxton task update TASK-001 --status in_progress
clauxton task update TASK-001 --priority critical
# Get next recommended task (AI-powered)
clauxton task next
# Delete task
clauxton task delete TASK-001
MCP Server (Phase 1 - Available Now!)
The Clauxton MCP Server provides full Knowledge Base and Task Management for Claude Code:
// .claude-plugin/mcp-servers.json
{
"mcpServers": {
"clauxton": {
"command": "python",
"args": ["-m", "clauxton.mcp.server"],
"cwd": "${workspaceFolder}"
}
}
}
Knowledge Base Tools:
kb_search(query, category?, limit?)- Search with TF-IDF relevance rankingkb_add(title, category, content, tags?)- Add new entrykb_list(category?)- List all entrieskb_get(entry_id)- Get entry by IDkb_update(entry_id, title?, content?, category?, tags?)- Update entrykb_delete(entry_id)- Delete entry
Note: Search results are automatically ranked by relevance using TF-IDF algorithm. Most relevant entries appear first.
Task Management Tools (✅ Week 5):
task_add(name, description?, priority?, depends_on?, files?, kb_refs?, estimate?)- Add tasktask_list(status?, priority?)- List tasks with filterstask_get(task_id)- Get task detailstask_update(task_id, status?, priority?, name?, description?)- Update tasktask_next()- Get AI-recommended next tasktask_delete(task_id)- Delete task
See MCP Server Guide for complete documentation.
Coming in Phase 2
# Conflict detection
/conflicts-check
Knowledge Base YAML Structure
After running clauxton kb add, your entries are stored in .clauxton/knowledge-base.yml:
version: '1.0'
project_name: my-project
entries:
- id: KB-20251019-001
title: Use FastAPI framework
category: architecture
content: |
All backend APIs use FastAPI framework.
Reasons:
- Async/await support
- Automatic OpenAPI docs
- Excellent performance
tags:
- backend
- api
- fastapi
created_at: '2025-10-19T10:30:00'
updated_at: '2025-10-19T10:30:00'
version: 1
Categories:
architecture: System design decisionsconstraint: Technical/business constraintsdecision: Important project decisions with rationalepattern: Coding patterns and best practicesconvention: Team conventions and code style
See YAML Format Reference for complete schema documentation.
🏗️ Architecture
Current (Phase 0-1)
clauxton/
├── core/
│ ├── models.py # Pydantic data models ✅
│ └── knowledge_base.py # KB CRUD operations ✅
├── utils/
│ ├── yaml_utils.py # Safe YAML I/O ✅
│ └── file_utils.py # Secure file operations ✅
├── cli/
│ └── main.py # CLI commands ✅
└── mcp/
└── server.py # MCP Server ✅ (Phase 1, Week 3)
Storage: .clauxton/knowledge-base.yml (YAML format)
Planned (Phase 1-2)
- Task Management MCP Tools: task_add, task_list, task_next (Week 4)
- Dependency Analysis: Auto-inference, DAG validation (Week 5-6)
- Enhanced Search: TF-IDF relevance (Week 7)
- Conflict Detection: Pre-merge conflict analysis (Phase 2)
See docs/architecture.md for complete design.
📚 Documentation
User Guides
- Quick Start Guide - Get started in 5 minutes (CLI)
- MCP Server Quick Start - Get started with Claude Code ✨ NEW
- Task Management Guide - Complete task management documentation ✨ NEW
- Search Algorithm - TF-IDF search explanation ✨ NEW
- Installation Guide - Complete installation instructions
- YAML Format Reference - Complete Knowledge Base YAML specification
- MCP Server Guide - Complete MCP Server documentation
Developer Guides
- Architecture Overview - System design and data flow
- Development Guide - Setup and contribution guide
- Technical Design - Implementation details
- Roadmap - 16-week development plan
- Contributing - Contribution guidelines
Coming Soon
- API Reference (Phase 1)
- Configuration Guide (Phase 1)
🤝 Contributing
We welcome contributions! Please see CONTRIBUTING.md for guidelines.
📄 License
MIT License - see LICENSE for details.
📊 Project Status
| Phase | Status | Completion | Target Date |
|---|---|---|---|
| Phase 0: Foundation | ✅ Complete | 100% | Week 2 (2025-11-02) |
| Phase 1: Core Engine | 🚧 In Progress | 88% | Week 3-8 |
| Phase 2: Conflict Prevention | 📋 Planned | 0% | Week 9-12 |
| Beta Testing | 📋 Planned | 0% | Week 13-14 |
| Public Launch | 📋 Planned | 0% | Week 15-16 |
Phase 0 Progress (Complete ✅):
- ✅ Pydantic data models (100%)
- ✅ YAML utilities (100%)
- ✅ Knowledge Base core (100%)
- ✅ CLI implementation (100%)
- ⏳ Basic MCP Server (0% - deferred to Phase 1)
- ✅ Tests & Documentation (100% - 111 tests, 93% coverage)
Phase 1 Progress (Week 9/10 - 95%):
- ✅ Knowledge Base CRUD (100% - complete with update/delete)
- ✅ MCP KB Tools (100% - kb_search, kb_add, kb_list, kb_get, kb_update, kb_delete)
- ✅ Task Management (100% - CRUD, dependencies, DAG validation, CLI)
- ✅ Task Management MCP Tools (100% - task_add, task_list, task_get, task_update, task_next, task_delete)
- ✅ Auto Dependency Inference (100% - file overlap detection)
- ✅ TF-IDF Search (100% - relevance-based search with fallback)
- 🚧 Documentation (80% - Week 9-10)
- ✅ Tests: 265 total, 94% coverage
See Phase 0 Completion Summary for detailed results. See docs/roadmap.md for overall timeline. See docs/phase-1-plan.md for next steps.
🔗 Links
- GitHub: https://github.com/nakishiyaman/clauxton
- Issues: https://github.com/nakishiyaman/clauxton/issues
- Discussions: https://github.com/nakishiyaman/clauxton/discussions
- PyPI: Coming after Phase 0 completion
🙏 Acknowledgments
This project was inspired by the need for persistent context in AI-assisted development. Special thanks to the Claude Code team for building an extensible platform.
Note: Clauxton is an independent project and is not officially affiliated with Anthropic or Claude Code.
Built with ❤️ for Claude Code users
Project details
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 clauxton-0.8.0.tar.gz.
File metadata
- Download URL: clauxton-0.8.0.tar.gz
- Upload date:
- Size: 171.2 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
2d7313fe4b1da0e2202de3fa5c681ddfea55d8fe8ce6229e64316d801e419535
|
|
| MD5 |
391f2f7468ca6210c2b0ac1eda6529ea
|
|
| BLAKE2b-256 |
39e6a75be5e2c82f350334cefa26c4754d691e9aab8e9d73d4010836d7cd67c3
|
File details
Details for the file clauxton-0.8.0-py3-none-any.whl.
File metadata
- Download URL: clauxton-0.8.0-py3-none-any.whl
- Upload date:
- Size: 33.8 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
72d4b5e70c94a9d9bda5c26df7651d83b03104a391745058b18f45fa77ed1fc8
|
|
| MD5 |
1b2a521a9aa94b7bbe229c3deb77ee39
|
|
| BLAKE2b-256 |
11df0413426451ebd20e60d7f71a56644ada15f0de123cad75bb82af8f194327
|