AI · 2025 · Kepler
KeplerOpsAgent
An autonomous ops agent that triages incidents, writes RCAs and files hotfixes.
Impact
Saved 24k engineer-hours
AgentsAutomationDevOps
Challenge
The problem we walked into.
Kepler’s SREs were drowning in pager duty — 8 minute MTTA, 2 hour MTTR, and burnout climbing.
Solution
What we built.
A LangGraph agent that reads logs, correlates traces, drafts hotfixes and opens PRs — with a human-in-the-loop for anything above P3.
Process
How we got there.
- 01Incident taxonomy + eval set
- 02Agent scaffolding + tool use
- 03Slack + PagerDuty integration
- 04Guardrails + kill switches
Technology
The stack.
LangGraphClaude 4DatadogPagerDutyGitHub Actions
Before → After
Drag to compare.
Before
After
Results
Impact you can measure.
78%
Tickets auto-resolved
–6m
MTTA improvement
24k
Engineer-hours saved
0
Regressions
