What tools do you use for monitoring multi-agent teams?

[avoronov-explyt]

What tools do you use for monitoring multi-agent teams?

We're running a 9-agent team in production (PO, BA, Design, SM, Backend, Frontend, QA, DevOps, Manager) and the monitoring gap is real.

Our current stack is... basic: each agent writes structured logs to shared markdown files, and a Manager agent aggregates them. It works, but we're flying blind on a lot of metrics.

What we want to track:

• Handoff failure rate — ~40% of agent handoffs need revision. This is our Create enum for session end states and event results #1 quality signal.
• Sprint velocity — are we shipping faster or slower this sprint? Currently just counting completed tasks.
• Stuck task detection — tasks in Review/In Progress for >2x median time.
• Token efficiency — what % of LLM tokens go to productive work vs. coordination overhead?
• Hallucination propagation — how often does a hallucination in Agent A reach Agent C before QA catches it?

What we track now (manually):

• Task status transitions on the Kanban board
• QA pass/fail at each handoff
• Overall sprint completion rate

We built 13 tools in one session and documented the gaps: How an AI Agent Team Built 13 Products in 4 Hours

Our coordination templates: Aries AI Tools

What's your monitoring setup? Any tools that work well for multi-agent observability?