Overview

Context is what Polylane sees. Detection is what it spots. Investigation is where an agent works the problem: it queries your observability providers, reads your code, walks the topology, and writes up what it found. Every tool call and result is stored in the transcript, and answers cite their sources, so you can check any number the agent puts in front of you.

Threads

Every agent run happens inside a thread: a persistent transcript of every message, tool call, and result. You start some by hand as chat; others are opened by automations, investigations, or autofixes. Inside a thread the agent produces artifacts: reports, diagrams, hypotheses, pull requests. See Threads.

Investigations

An investigation is the coordinating agent engagement on one confirmed issue, embedded in the issue itself. It spawns hypothesis sub-agents that each test a named hypothesis and return a verdict: confirmed, refuted, or inconclusive, with a confidence level. Everything lands on the issue's timeline. See Investigations.

Escalations

When an investigation parks needing a human, the ask is recorded as an escalation: one record per distinct ask, however often it recurs, with one notification. Clear it when handled. See Escalations.

Where investigations come from

• You ask a question in a chat thread.
• Triage confirms an issue as an incident and its investigation starts automatically.
• An automation run opens an issue with the openIssue action.
• Change intelligence opens a review thread for a pull request.

When to reach for which

Ask an agent to start one now:

Use polylane to start a thread and investigate: why did error rates on our api-gateway service change in the last 24 hours? Cite every number back to its source.