Incident memory & living runbooks
“Last time Redis did this, we…” — outages become recallable procedure.
Without memory
At 3am, the person who fixed this exact outage last year is asleep, the postmortem is in a doc nobody can find, and the runbook is eighteen months stale. Teams pay for the same incident twice because the fix lived in someone’s head.
With Rekall
Every incident becomes memory: what broke, what was tried, what worked, and why. Successful mitigations codify into executable runbooks, and the on-call’s AI recalls the relevant history the moment symptoms match.
How it works
Incidents are captured as they happen
The debugging session — hypotheses, dead ends, the fix — lands in episodic memory with the decision trail attached.
Fixes codify into runbooks
Workflow detection turns repeated successful mitigations into structured, executable skills — living runbooks that update themselves with each incident.
On-call recalls instantly
“Redis latency spiking after deploy” surfaces the two previous episodes, what worked, and the runbook — at 3am, without waking anyone.
See it in action
> redis latency spiking after the 14:00 deployrekall.recall("redis latency spike", hive: "platform")→ 2 prior incidents match:INC-114 (Feb): fragmentation after BGSAVE —fixed via activedefrag; runbook codified ✓INC-201 (May): noisy-neighbour on shared VPS —ruled out in 10 min via slowlog checkSuggested runbook: redis-defrag.skill (lastverified 3 weeks ago)
Related use cases
Skill codification — workflows that teach themselves
Work out a tricky procedure once. Rekall turns it into a skill your whole team runs.
The “why” archive — decision memory
“Why is auth done this weird way?” finally has an answer.
Compliance-grade audit memory
Who decided what, when, on what evidence — and what you believed at the time.