Mechifact ⚡

Multi-Agent System Observability Platform 🔍

Gain complete visibility into complex multi-agent systems. Monitor every tool call, trace agent interactions, and debug distributed workflows with comprehensive observability designed for modern AI architectures.

Join Waitlist

The Problem

Complex agentic systems are difficult to debug, measure, and improve. As AI agents become more sophisticated and work together in distributed architectures, understanding what's happening under the hood becomes increasingly challenging:

• Tool calls happen across multiple agents with no unified view
• Agent coordination is opaque and hard to trace
• System debugging requires piecing together logs from different sources
• Performance bottlenecks are hidden in complex interaction patterns

The Solution

Mechifact provides comprehensive observability for multi-agent systems through three core capabilities:

🔧 Tool Call Tracing

Monitor every tool invocation across your multi-agent system with detailed execution traces and performance metrics.

🤝 Agent Coordination

Visualize how agents communicate, coordinate, and share information across your distributed system architecture.

🐛 System Debugging

Debug complex multi-agent workflows with comprehensive logs, error tracking, and performance bottleneck identification.

Quick Start

Prerequisites

• Python 3.8+
• OpenAI API key or compatible LLM provider

Installation

1. Clone the repository

   git clone https://github.com/ltejedor/mechifact.git
   cd mechifact

2. Install dependencies

   pip install -r requirements.txt
   pip install flask python-dotenv markdownify  # for live server
   pip install networkx pyvis  # for visualizations

3. Set up environment

   cp .env.example .env
   # Edit .env with your LLM API keys

Run Live Observability Server

Start the real-time monitoring interface:

python live_server.py

Open http://localhost:8000 in your browser and start giving tasks to your agent. Watch as tool calls, agent interactions, and system state changes appear in real-time.

Generate Static Visualizations

Create knowledge graph visualizations from agent execution logs:

# Generate knowledge graph
python 01-knowledge-graph/viz_knowledge_graph.py output/proof_of_work_*.json

# Generate combined timeline + graph view
python 02-timeline/viz_multi_view.py output/proof_of_work_*.json

Features

• Real-time Monitoring: Live web interface showing agent steps as they happen
• Knowledge Graph: Interactive visualization of agent relationships and data flow
• Timeline View: Chronological trace of all tool calls and agent actions
• Multi-View Dashboard: Combined graph, timeline, and detailed logs
• Framework Agnostic: Built on smolagents with extensible architecture

Architecture

Mechifact works by:

1. Instrumenting your multi-agent system to capture provenance data
2. Streaming real-time events to the observability dashboard
3. Visualizing agent interactions through multiple complementary views
4. Analyzing performance patterns and debugging information