LLM Session Manager is a free, open-source monitoring and collaboration platform for AI coding assistants. It provides unified monitoring across Claude Code, Cursor, and GitHub Copilot with health scoring, token tracking, and team analytics.
- ✅ Claude Code (Anthropic)
- ✅ Cursor IDE
- ✅ GitHub Copilot
- 🔜 More coming soon!
Yes! 100% free and open-source (MIT License). No hidden costs, no paid tiers, no limitations.
- ✅ macOS
- ✅ Linux
- ✅ Windows (WSL recommended)
100% locally on your machine. Your code and session data never leave your computer unless you explicitly use the team collaboration features.
No. LLM Session Manager only monitors your AI coding sessions locally. It doesn't access your code or send data to external servers.
Yes! Since everything runs locally, it's safe for corporate/enterprise use. Your company's code stays private.
| Feature | Claude Analytics | LLM Session Manager |
|---|---|---|
| Price | $30-60/user/month | Free |
| Multi-tool | Claude only | Claude + Cursor + Copilot |
| Health scoring | No | Yes |
| Local/private | Cloud-based | 100% local |
| Team collaboration | Limited | Full-featured |
Completely different use cases:
- AWS AgentCore: Production AI agent infrastructure (for deploying customer-facing agents)
- LLM Session Manager: Development monitoring (for developers using AI coding tools)
See AWS AgentCore Comparison for details.
Most AI coding tools don't have built-in analytics, and if they do:
- ❌ They're tool-specific (can't see Cursor + Copilot together)
- ❌ They require expensive enterprise plans
- ❌ They don't have health scoring
- ❌ They don't warn you before sessions degrade
- Python 3.10+
- 2GB RAM minimum
- 500MB disk space
- Active AI coding assistant (Claude/Cursor/Copilot)
The tool is designed to monitor AI coding sessions. Without an active Claude Code, Cursor, or Copilot session, there's nothing to monitor.
LLM Session Manager uses a hybrid detection system:
- Process inspection (finds running AI tool processes)
- Registry-based detection (checks known AI tool signatures)
- Zero configuration required - it just works!
No. LLM Session Manager uses minimal resources (< 1% CPU) and doesn't interfere with your AI tools.
The health score tells you! When it drops below 70%, you'll get recommendations:
- 🟢 90-100: Session is healthy
- 🟡 70-89: Minor issues, monitor closely
- 🟠 40-69: Consider restarting soon
- 🔴 0-39: Restart recommended
Yes! That's the point. Track Claude Code, Cursor, and Copilot all in one dashboard.
Token counts are estimated using tiktoken (same library Claude uses). Accuracy: ~95-98%.
Yes! Export to:
- JSON (for integrations)
- YAML (for configs)
- Markdown (for reports)
- Start the collaboration backend server
- Share a session using
llm-session share <session-id> - Teammates can view sessions via web dashboard
- Real-time updates via WebSocket
Yes! Self-hosted collaboration is 100% free.
For viewing: Just a web browser (access the dashboard URL) For full features: Install LLM Session Manager
- Make sure you have Claude Code, Cursor, or Copilot running
- Try running:
llm-session discover - Check if processes are running:
ps aux | grep -i "claude\|cursor\|copilot"
Run llm-session list first - this discovers and saves sessions to the database.
Health scores are based on:
- Token usage (higher usage = lower score)
- Session duration (very long = lower score)
- Activity level (idle = lower score)
- Error count (more errors = lower score)
Make sure you installed with: poetry install
If issues persist: poetry install --no-cache
See CONTRIBUTING.md for:
- Setting up development environment
- Running tests
- Code style guidelines
- Pull request process
GitHub Issues: https://github.com/iamgagan/llm-session-manager/issues
Absolutely! Open a GitHub Issue with the "feature request" label.
- 🔜 VS Code extension
- 🔜 Web dashboard (alternative to CLI)
- 🔜 Slack/Discord notifications
- 🔜 Custom health metrics
- 🔜 More AI tool integrations
Check our GitHub Projects for roadmap and timelines.
- 📖 Read the User Guide
- 💬 Open a GitHub Discussion
- 🐛 Report a GitHub Issue
- ⭐ Star us on GitHub