Complete Visibility Into
AI Agent Behavior
OBSERVE
Every Action
AUDIT
Every Decision
REPORT
Every Insight
Framework Integration Hub
Works with MCP, LangChain, CrewAI, and custom agents - everything you need for universal observability
Complete MCP Documentation
Comprehensive guides for integrating AOP with Model Context Protocol (MCP) tools and servers.
- Tool execution observability with decorators
- LLM sampling request/response tracking
- Automatic parameter and result capture
- Error handling and exception tracking
@client.mcp.observe_tool(agent_id='my-agent')
def search_tool(query: str, max_results: int = 10):
"""Search for information."""
results = perform_search(query, max_results)
return {'results': results, 'count': len(results)}See It In Action
Real-time observability for every agent action, decision, and insight
Coming Soon
Demo Video
Watch a complete walkthrough of AOP in action
Installation → Integration → Real-time Insights
Live Event Stream
| Timestamp | Agent | Event | Duration | Status |
|---|---|---|---|---|
| 10:23:45 | search-agent | mcp.tool.called | 245ms | |
| 10:23:42 | orchestrator | a2a.task.assigned | 12ms | |
| 10:23:40 | search-agent | mcp.sampling.request | 1,234ms | |
| 10:23:38 | worker-agent | mcp.tool.called | 89ms | |
| 10:23:35 | payment-agent | ap2.payment.initiated | 342ms |
Powerful Features
Everything you need for complete observability
Dashboard Table View
Professional tabular interface with sortable columns, live updates, and click-to-view details.
- Sort by timestamp, agent, event type, or duration
- Color-coded status indicators
- Real-time WebSocket streaming
- Detailed JSON viewer on click
Trace Explorer - 3 Search Methods
Interactive tree view of distributed traces with multiple search methods: Correlation ID, Event ID, or Parent ID.
- Search by Correlation ID for planned workflows
- Search by Event ID - no correlation ID needed!
- Search by Parent ID for sub-operations
- Complete trace reconstruction with parent-child relationships
Analytics Charts
Real-time performance metrics, aggregations, and time-series analysis.
- Tool usage statistics
- Latency percentiles (P95, P99)
- Event rate monitoring
- Time-bucketed timelines
CLI & Export Tools
Powerful command-line interface for querying, exporting to 5 formats, and monitoring.
- Query events with rich filters
- Export to JSON, CSV, TOON (30-60% token savings)
- OpenTelemetry and Prometheus export
- Interactive trace viewer and analytics
Why AOP?
Designed for production with privacy, performance, and simplicity in mind
P99 Latency
Minimal overhead, production-ready performance
Protocols
MCP, A2A, and AP2 support out of the box
Privacy
Local storage by default, you own your data
Dependencies
Core library uses only Python stdlib
Join developers building transparent, auditable AI systems
What's Next
Building the future of agent observability with compliance and enhanced protocols
HIPAA Compliance Toolkit
Healthcare-grade data protection and audit trails for medical AI agents
GDPR Compliance Toolkit
European privacy standards with right-to-deletion and consent tracking
AI Agentic Tool Call & MCP
Pre-built observability for AI agent tool calls and MCP server integration
A2A Protocol
Agent-to-Agent communication tracking for multi-agent workflows
AP2 Protocol
Agent Payments Protocol for tracking costs and transactions
Coming Soon
Stream processing, batch optimization, and more storage backends
Want to shape the roadmap? We're listening to the community.
Join the DiscussionGet in Touch
Questions, feedback, or collaboration? Reach out to the developer.
Ajit Singh
Creator & Maintainer
Building transparent AI systems
Open to collaborations, contributions, and feature requests