🔍 Make AI Agents Transparent & Auditable
Capture, track, and audit every AI agent action
OpenClaw Transparency Layer is a lightweight, open-source solution for capturing and auditing AI agent sessions. Inspired by Entire Checkpoints, but designed specifically for the OpenClaw ecosystem.
- 🔍 Auditability: Know exactly what your AI agents did and why
- 🐛 Debugging: Quickly identify where agents went wrong
- 📊 Compliance: Generate audit trails for enterprise requirements
- 📚 Onboarding: Help new team members understand AI workflows
- 🔄 Reproducibility: Recreate agent sessions from any checkpoint
v0.1.0 - Single Agent (MVP)
- ✅ Session Recording: Capture every AI agent action (prompts, tool calls, decisions)
- ✅ Checkpoint System: Create save points at any time
- ✅ Auto-Summarization: Automatically generate session summaries
- ✅ Lightweight Storage: File-based, no database required
- ✅ Easy Integration: Add transparency to any OpenClaw agent in 2 lines of code
v0.2.0 - Multi-Agent & HTML Visualization (NEW! 🚀)
- ✅ Multi-Agent Tracking: Track interactions between multiple agents
- ✅ Interaction Types: Delegation, request, response, collaboration, handoff
- ✅ Coordination Tracking: Monitor agent coordination and orchestration
- ✅ Conflict Detection: Automatically detect resource contention and conflicts
- ✅ HTML Visualization: Beautiful, interactive HTML reports with Mermaid.js charts
- ✅ Responsive Design: Mobile-friendly reports with gradient colors
- ✅ Timeline Visualization: Visual timeline of agent interactions
- ✅ Statistics Dashboard: At-a-glance metrics with stat cards
- ✅ 4+ Concurrent Agents: Support for complex multi-agent systems
- 🔜 Enterprise Compliance: HIPAA, SOX, GDPR audit reports
- 🔜 Real-time Dashboard: Live monitoring with WebSocket updates
- 🔜 Git Integration: Automatic commits on checkpoints
- 🔜 PDF Export: Generate PDF reports from sessions
- 🚀 Cloud Sync: Sync sessions across team members
- 🚀 AI-Powered Insights: Get recommendations from session analysis
- 🚀 Enterprise Features: SSO, audit logs, compliance reports
pip install openclaw-transparencyOr use the standalone module:
# Download the single-file MVP
curl -O https://raw.githubusercontent.com/your-username/openclaw-transparency/main/transparency.pyfrom openclaw import Agent
from openclaw_transparency import TransparencyLayer
# Initialize your OpenClaw agent
agent = Agent("my-agent")
# Add transparency in 2 lines
transparency = TransparencyLayer(agent_name="my-agent")
transparency.enable()
# Now all agent actions are automatically tracked
result = agent.execute("Create a REST API endpoint")
# Create checkpoints at key moments
transparency.create_checkpoint(
description="Completed API implementation",
files_modified=["api.py", "test_api.py"]
)
# End session and get summary
summary = transparency.end_session()✅ Transparency Layer enabled for agent: my-agent
📋 Session ID: 2026-02-16-abc123
📝 Tracked action: prompt
📝 Tracked action: tool_call
📝 Tracked action: decision
✅ Checkpoint created: a3b2c4d5e6f7
📊 Session Summary:
{
"session_id": "2026-02-16-abc123",
"duration_seconds": 45.2,
"total_actions": 12,
"total_checkpoints": 3,
"files_modified": ["api.py", "test_api.py", "README.md"],
"key_decisions": ["Use FastAPI framework", "Add JWT authentication"]
}
from multi_agent_transparency import MultiAgentTransparency
# Initialize multi-agent system
multi_agent = MultiAgentTransparency(project_name="Web Application")
# Register agents
multi_agent.register_agent(
agent_name="CodeGenerator",
agent_type="developer",
capabilities=["code_generation", "refactoring"]
)
multi_agent.register_agent(
agent_name="Reviewer",
agent_type="reviewer",
capabilities=["code_review", "quality_check"]
)
# Track agent interactions
multi_agent.track_interaction(
from_agent="CodeGenerator",
to_agent="Reviewer",
interaction_type="delegation",
content="Please review the authentication module"
)
# Generate HTML report
html_content = multi_agent.generate_html_report()
# End session
summary = multi_agent.end_session()HTML Report Features:
- 📊 Interactive Graph: Mermaid.js visualization of agent workflows
- ⏱️ Timeline View: Visual timeline of all interactions
- 🤖 Agent Cards: Beautiful cards showing agent capabilities
- 📈 Statistics: At-a-glance metrics (agents, interactions, conflicts)
- 🎨 Responsive Design: Works on desktop and mobile
- 🌈 Modern UI: Gradient colors, hover effects, smooth transitions
Open the generated HTML file in any browser to see the interactive report!
Initialize transparency layer for an agent.
Parameters:
agent_name(str): Name of your AI agentstorage_path(str, optional): Where to store session data (default:./transparency-sessions)auto_save(bool, optional): Auto-save after each action (default:True)
Manually track an agent action.
Parameters:
action_type(str): Type of action (e.g., "prompt", "tool_call", "decision")input_data(any): Input to the actionoutput_data(any): Output from the actionmetadata(dict, optional): Additional context
Create a checkpoint (save point) in the session.
Parameters:
description(str): Human-readable descriptionfiles_modified(list, optional): List of files modifieddecisions(list, optional): List of key decisions
End the session and generate final summary.
Returns:
dict: Session summary with statistics and key insights
- Basic session recording
- Checkpoint system
- Auto-summarization
- File-based storage
- Multi-agent interaction tracking
- Simple visualization dashboard
- Git integration
- Better error handling
- Team collaboration features
- Export to PDF/JSON
- Performance optimizations
- Extended documentation
- Cloud sync
- AI-powered insights
- Enterprise features
- Plugin system
# Track your personal AI coding assistant
transparency = TransparencyLayer("my-coding-assistant")
# Review sessions later to improve prompts# Share session logs with team members
# Understand why certain decisions were made
# Onboard new developers faster# Generate compliance reports
# Audit AI usage across organization
# Meet regulatory requirements (SOC 2, HIPAA)We welcome contributions! This is an open-source project aimed at making AI agents more transparent and trustworthy.
Ways to contribute:
- 🐛 Report bugs
- 💡 Suggest features
- 📖 Improve documentation
- 🔧 Submit pull requests
See CONTRIBUTING.md for details.
| Feature | OpenClaw Transparency | Entire Checkpoints | Git LFS |
|---|---|---|---|
| Open Source | ✅ MIT | ✅ MIT | ✅ |
| Agent Agnostic | ✅ | ❌ Claude/Gemini only | ✅ |
| Multi-Agent | 🔜 v0.2.0 | ❌ | ❌ |
| Git Integration | 🔜 v0.2.0 | ✅ | ✅ |
| Visualization | 🔜 v0.2.0 | ❌ | ❌ |
| Easy Setup | ✅ 2 lines | ||
| File-Based | ✅ | ❌ Shadow branch | ✅ |
Our Differentiation:
- Focus on OpenClaw ecosystem
- Agent-agnostic design
- Multi-agent tracking (coming soon)
- Easier setup (no Git hooks required for basic usage)
- ✅ Full session recording
- ✅ Checkpoint system
- ✅ Auto-summarization
- ✅ File-based storage
- ✅ Community support
- ✅ Everything in Open Source
- ✅ Multi-agent tracking
- ✅ Visualization dashboard
- ✅ Export to PDF/JSON
- ✅ Priority support
- ✅ Everything in Pro
- ✅ Cloud sync
- ✅ Team collaboration
- ✅ Compliance reports
- ✅ SSO integration
- ✅ Dedicated support
- Inspired by Entire Checkpoints by Thomas Dohmke
- Built for the OpenClaw ecosystem
- Thanks to the open-source community
MIT License - see LICENSE for details.
- GitHub Issues: Report bugs or request features
- Discord: Join our community
- Twitter: @OpenClawAI
⭐ If this project helps you, please give it a star! ⭐
Made with ❤️ for the OpenClaw community