Devin launches Auto-Triage with long-term memory for bugs, alerts, and incidents

• cognition's launch thread positions Devin Auto-Triage as an always-on incident and bug investigator that watches incoming bugs, alerts, and incidents, then returns context, next steps, or a PR.
• According to cognition's product thread, the system plugs into Slack, Linear, GitHub, schedules, webhooks, observability tools, tickets, alerts, and code, which makes this a workflow product as much as a coding agent feature.
• The differentiator Cognition pushed hardest in cognition's product thread is long-term memory: Devin can remember prior investigations, deduplicate related reports, link threads and tickets, map ownership areas, and track recurring issues.
• Early public usage already extends beyond a canned demo, with cognition's Modal quote saying Modal is using it on an inference team and benhylak's Raindrop post showing it pointed at Raindrop incidents.
• russelljkaplan's post framed the bigger product bet clearly: prompting an agent by hand is turning into the fallback path, while scheduled and event-driven automations become the default entry point.

You can browse the official Auto-Triage page, jump straight to Cognition's launch thread, and inspect a separate automation gallery example that kicks off PR work after a Granola call ends. One of the more useful details is that the launch is not limited to Slack intake, because Cognition's own materials and benhylak's screenshot tie it to Datadog, Sentry, PagerDuty, and Raindrop. The weirdest tell is probably russelljkaplan's wording: manual prompting as a UX bug.

Cognition shipped Auto-Triage as a front door for Devin, not a one-off incident bot. The product claim in cognition's launch thread is simple: Devin monitors incoming issues, starts investigating on its own, and can come back with either context, next steps, or a code change.

The launch video attached to cognition's launch thread shows the intended loop: a Slack alert lands, Devin investigates, and the flow ends with a GitHub issue containing the write-up. That puts the product closer to automated on-call intake than to a chat-first coding assistant.

Russell Kaplan, president at Cognition, made the thesis explicit in russelljkaplan's launch-day post: the company thinks persistent automations will replace a lot of explicit prompting. That is the real product direction here.

The most concrete new mechanism in the launch is memory. In cognition's product thread, Cognition says Auto-Triage can:

• remember prior investigations
• deduplicate related reports
• link threads and tickets
• map team ownership areas
• keep track of recurring issues

That list matters because most alert bots stop at summarization. As dabit3's repost of walden_yan put it, the interesting part is that Devin is being framed more like a manager agent coordinating investigation work than a brittle if-this-then-that automation.

Cognition's own thread says Auto-Triage connects to Slack, Linear, GitHub, schedules, webhooks, observability tools, tickets, alerts, and code. benhylak's post adds the clearest ecosystem snapshot, with a screenshot showing Datadog, Sentry, PagerDuty, and Raindrop under a "Works with your existing tools" banner.

That breadth is why the product reads as incident plumbing rather than a single chat integration. mattlam_'s reaction argued that other coding agents already have Slack hooks, but this is the first announced product he had seen built specifically around the on-call and auto-triage problem.

Cognition launched with at least one named production reference. In cognition's Modal quote, Hari Subbaraj, Member of Technical Staff at Modal, said Devin Auto-Triage is already being used for incidents on Modal's inference team and described it as more useful than previous auto-triage tools because it works across the codebase and observability stack.

The public evidence is still thin, but it is not just vendor copy. Modal replied positively in modal reply, and benhylak's Raindrop post claimed Devin could already auto-triage Raindrop issues.

Auto-Triage shipped alongside a broader automation story. In imjaredz's post, a Cognition employee described using Devin to kick off PRs automatically after each Granola call, then linked the public Granola post-call processing guide.

That makes the launch feel like one surface of a larger event-driven Devin platform. Cognition also attached a limited-time onboarding carrot to the release, with cognition's product thread offering $200 in credits for building a first automation.