Put a swarm of agents on every hard incident.
A Polylane swarm sends a team of agents after every possible cause of an incident at once, checks their findings against each other, and hands you the answer.
No credit card required.
Get the root cause before everyone joins the incident call.
Hard incidents rarely have one obvious cause. A swarm checks the database, the last deploy and the edge at the same time, so you're not waiting on each theory in turn.
A second agent checks every finding before you see it.
Every finding is checked by a second agent before it reaches you. The ones that don't hold up get crossed out.
You get the cause, the evidence and the fix in one message.
When the swarm is done, you get the cause, the evidence behind it and the fix, in one message.
Checkout is throwing 502s. Can you put a swarm on it?
Deploy 9f3c2a1 cut the hd-prod connection pool from 50 to 5, so checkout requests queue for a connection and time out. Two agents confirmed it; the edge and traffic theories were ruled out.
The swarm adds agents when it finds new leads, up to 25.
It starts with a few agents. When the lead finds more leads worth chasing, it adds agents, up to 25.
Outcomes
Real numbers from Polylane customers.
Common questions.
When should I use a swarm?
When an incident has more than one possible cause, or when you need the answer fast. Pick Swarm when you start a thread, or let an investigation grow into one.
How big can a swarm get?
Up to 25 agents. It starts small and adds agents as the work grows.
Which plans include swarms?
Every paid plan. You also need GitHub connected.
Can a swarm change production?
Only the lead agent can propose a change, and it goes through the same approval as any other. The other agents read and test.
Can I see what each agent is doing?
Yes. The thread lists every agent in the swarm with what it is working on, and the shared board shows their findings.