name: trinity-connect description: Listen for real-time events from Trinity agents using WebSocket connection. allowed-tools: [Bash, Read] user-invocable: true argument-hint: "[agent-name] [state-filter]" automation: manual

Trinity Connect - Real-time Event Listener

Listen for real-time events from Trinity agents using WebSocket connection.

State Dependencies

Prerequisites

1. MCP API Key: Get from Trinity Settings → API Keys
2. websocat: Install with brew install websocat (or wscat via npm)
3. jq: Install with brew install jq

Usage

Set your API key and run the listener:

export TRINITY_API_KEY="trinity_mcp_xxx"
./scripts/trinity-listen.sh [agent-name] [state-filter]

Arguments

Examples

# Listen for all events from all accessible agents
./scripts/trinity-listen.sh

# Listen for events from a specific agent
./scripts/trinity-listen.sh my-research-agent

# Listen for completed events only
./scripts/trinity-listen.sh all completed

# Listen for when a specific agent completes work
./scripts/trinity-listen.sh my-agent completed

Event Types

Event-Driven Loop Pattern

Use this pattern for local-remote agent coordination:

while true; do
    echo "Waiting for agent to complete..."
    EVENT=$(./scripts/trinity-listen.sh my-agent completed)

    # Parse the event
    AGENT=$(echo "$EVENT" | grep -A20 "TRINITY EVENT" | jq -r '.name // .agent_name')

    # React to the event
    echo "Agent $AGENT completed! Processing results..."
    # ... your processing logic here ...
done

Troubleshooting

Related

• Feature docs: docs/memory/feature-flows/trinity-connect.md
• MCP API Keys: Settings → API Keys in Trinity UI

Completion Checklist

• API key configured
• WebSocket tools installed
• Connection established
• Events received and parsed
• Filters applied correctly