Anthropic demos Claude as on-call engineer: 15-minute incident triage in Slack
xiaohu · x · 2026-09-14
Anthropic posted a video showing an engineer using Claude ("Claude Tag") in Slack to handle a night shift:
- At 11pm, a payments API error rate spikes past 2% and pages the on-call engineer — Claude is already investigating before she opens her laptop
- It pulls monitoring data, compares recent deploys, and reviews feature-flag changes, locating the cause in 15 minutes: a new feature's retry logic had no concurrency cap and flooded the queue
- It proposes two fixes, recommends one, and @-mentions the engineer for approval; only after she replies "add the cap, keep the feature on" does it change code — 4 lines total
- Post-deploy, it monitors the error rate for 10 minutes before declaring the incident resolved, and leaves a note: "check feature flags first"
The key design point is permission boundaries: it can investigate and propose code changes autonomously, but merging and deploying require human approval.
More from coding & agent
- Multi-agent harnesses vs one strong agent: builder finds the simpler setup wins — duku-95 · 2026-09-14
- OpenAI's managed Agents API raises the auditability question: what must stay inspectable? — Mediocre-Main-9870 · 2026-09-14
- Learning LangChain and CrewAI reveals there's more to AI engineering than agents — syed_kaif777 · 2026-09-14
- Agent got rescued by hand? Route that fix back into skill versioning — Jimcy-Maffesoli · 2026-09-14
- Property manager: Meta's Muse agent handled a 2am sewage backup end-to-end — armand_ruiz · 2026-09-14
- Memanto: open-source memory agent managing AI agents' memory, 2.2k stars — Shruti_0810 · 2026-09-14