Streams live swarm events as NDJSON for real-time agent observability.
npx skills add https://github.com/ruvnet/claude-flow --skill monitor-stream
Please help me install this Agent Skill. Skill: monitor-stream Source: https://github.com/ruvnet/claude-flow/tree/main/plugins/ruflo-swarm/skills/monitor-stream Command: npx skills add https://github.com/ruvnet/claude-flow --skill monitor-stream
Polling swarm status in a loop wastes tokens and misses transient events. This Skill streams swarm events in real time so you get immediate notifications when agents spawn, tasks complete, memory is written, or health checks run.
npx @claude-flow/cli@latest swarm watch --stream via the Monitor tool, emitting NDJSON events for agent spawns, task completions, memory writes, and health checks.swarm_status and swarm_health when a single snapshot is sufficient.swarm status.Ask the assistant to monitor the swarm in real time using the monitor-stream skill so you receive a notification for every agent spawn and task completion.
High-intent search queries and answers about installing and using this skill.
Run `npx @claude-flow/cli@latest swarm watch --stream` through the Monitor tool. It streams NDJSON events for agent spawns, task completions, memory writes, and health checks, with each stdout line triggering a notification.
Streaming is preferred over polling `swarm status` in a loop. The stream delivers events as they happen without repeated command invocations, as recommended in ADR-091. Use polling only for occasional one-shot checks.
Use the MCP tools `mcp__plugin_ruflo-core_ruflo__swarm_status` or `mcp__plugin_ruflo-core_ruflo__swarm_health` for a single status or health snapshot instead of starting a continuous event stream.
The stream emits NDJSON events covering agent spawns, task completions, memory writes, and health checks. Each line of stdout from the watch command triggers a separate notification through the Monitor tool.
Avoid the stream when you only need a single point-in-time check, since a long-running stream is unnecessary overhead. In that case call the swarm_status or swarm_health MCP tools once instead.
rafaelcallejaStream Claude Code agent events to a filesystem-based observability dashboard.
EmasoftTransfer critical work to an available agent during failures or blockers.
openeuler-mirrorCorrelate kernel logs, process memory snapshots, and cgroup metrics to diagnose Linux OOM events.
shuremali02Build a FastAPI chat backend with SSE streaming and SQLModel persistence.