Back to use cases
Engineering & SRE

Reconstruct yesterday's incident

PagerDuty + #incidents + meeting transcript + git history, woven into one timeline.

Best for
Post-mortem prep
Primitives
InteractionsEventsTime
Token savings
60%

The prompt

You are reconstructing the timeline of incident {{incident_id_or_date}}. Using Sentra: 1. Pull relevant data within the incident window (15 minutes before first page → 30 minutes after resolution) across PagerDuty, the #incidents Slack channel, any meeting that ran during the window, the post-mortem doc draft, and git push / deploy events. 2. Order strictly by timestamp. 3. For each row, label: source, actor, what happened in one sentence, and (if applicable) the Decision that was made at that moment. Output: - A timeline table: time · source · actor · event · decision (if any). - Three "key moments" with the quoted Interaction excerpt. - A first-draft Root Cause Hypothesis paragraph derived from the converged Rationale Sentra extracted from the responders.

See it work

Run it from

ClaudeCursorChatGPTAny agent · MCP

How it runs

  1. 01Copy the prompt into any agent connected to Sentra over MCP.
  2. 02The agent reads the company memory, scoped to what you are allowed to see.
  3. 03One grounded answer comes back, with every line traceable to its source.

Or automate it with Actions

The same prompt runs as a Sentra Action: on a schedule, or on a semantic trigger. Output lands in Slack or email, and every send waits for your approval.

Trigger idea: post-mortem prep

Sentraover MCPPagerDuty + #incidents + transcript + git, one timeline

Sep 3 checkout incident · reconstructed

02:14 · page

PagerDuty

p95 checkout latency crossed 4s. Jonas acknowledged in six minutes.

02:31 · rollback failed

#incidents · 02:31

The rollback hit the migration lock from Tuesday's deploy and aborted.

02:58 · cause identified

#incidents · 02:58

Priya traced it to the connection-pool change interacting with the A17 auth regression.

03:22 · mitigated

standup · Sep 3

Pool cap reverted; latency recovered by 03:40. Follow-up owner: Devon, revert-discipline note added to the runbook.

The prompt above becomes this. Fictional data, real mechanics: every line cites the meeting, thread, or record it came from.

Without Sentra

  1. 01Pull PagerDuty events in the incident window. Parse alert text and acknowledgments.
  2. 02Pull the #incidents Slack channel history for the window. Read every message to identify the material ones.
  3. 03Identify the meeting that ran during the incident — calendar lookup, transcript pull.
  4. 04Pull the post-mortem doc draft if it exists.
  5. 05Pull git push and deploy events from CI logs.
  6. 06Manually order every event by timestamp. Label source, actor, what happened.
  7. 07Re-read responder threads to extract converged root-cause hypothesis.

~75,000 tokens

With Sentra

  1. 01Pull every `Interaction` in the incident window across PagerDuty, Slack, meetings, docs, and git.
  2. 02Already ordered by timestamp, with source, `Actor`, and event content typed.
  3. 03`Decision` primitives extracted from responder threads — root cause hypothesis is structured.
  4. 04Compose the timeline table and the first-draft Root Cause Hypothesis.

~30,000 tokens

Book a DemoStart Free