How do I launch the observability dashboard?

You can activate the dashboard by telling Claude to 'start observability' or 'monitor agents', which launches the server on port 4000 and the UI on port 5172.

What kind of data does this skill monitor?

It tracks agent session starts/ends, tool calls, hook event executions, and provides a real-time timeline of all structured event data.

Does it store my agent history permanently?

No, the dashboard is designed to be ephemeral. It uses in-memory streaming and starts fresh each time you launch it.

What should I do if no events are showing in the UI?

Verify that your PAI hooks are configured to log to the correct JSONL file in your Claude history directory and that the 'capture-all-events' hook is active.

Agent Observability Dashboard

Name: Agent Observability Dashboard
Author: matt-fell-ai

bymatt-fell-ai

Analytics & Monitoring

Monitors PAI multi-agent activity in real-time through a live WebSocket-powered dashboard for debugging and workflow visualization.

About

The Agent Observability Dashboard provides deep visibility into the PAI multi-agent framework, allowing developers to track agent interactions, tool calls, and hook executions as they happen. By streaming logs directly from the filesystem to a Vite-based frontend via WebSockets, it offers an ephemeral, zero-database solution for identifying bottlenecks, debugging stuck workflows, and auditing complex agent behaviors in a clean, visual timeline.

Key Features

Automated tail-following of JSONL event logs for instant updates
Built-in management scripts for starting, stopping, and status checks
Real-time WebSocket streaming of agent sessions and tool calls
Live visual timeline for tracking multi-agent event sequences
Ephemeral in-memory architecture with no database requirements
0 GitHub stars

Use Cases

Auditing real-time agent activity to ensure proper hook integration
Debugging looping or stuck agents by monitoring live tool execution
Visualizing the sequence of events in complex multi-agent workflows

About

Key Features

Automated tail-following of JSONL event logs for instant updates
Built-in management scripts for starting, stopping, and status checks
Real-time WebSocket streaming of agent sessions and tool calls
Live visual timeline for tracking multi-agent event sequences
Ephemeral in-memory architecture with no database requirements
0 GitHub stars

Use Cases

Auditing real-time agent activity to ensure proper hook integration
Debugging looping or stuck agents by monitoring live tool execution
Visualizing the sequence of events in complex multi-agent workflows