Here is what we shipped to customers during the week of May 18, 2026: 4 product areas advanced across 5 backend and frontend repos, with 28 customer PRs landing in Annie backend v0.87.3, frontend v0.82.4, annie-intelligence v0.96.1, and agent-gateway v0.22.6, continuing the Anyshift Product Updates series from last week's post.

Personas for proactive reports, plus per-conversation effort levels

Annie gained two controls customers asked for. The proactive reports page now carries a persona setting, so the same project can adapt its report to whoever reads it (a dev team, an SRE rotation, a group of juniors) without standing up two pipelines. The chat header carries the same persona menu, so a one-off conversation can override the report default. Alongside it, an effort level shipped to both the project settings and the chatbar, controlling Annie's performance vs speed trade-off on a per-investigation basis: higher effort spends more reasoning budget per hypothesis. The pipeline resolves the chosen persona at the start of the report run, so audience and depth become first-class inputs rather than copy-paste edits after the fact.

OpsGenie joins the incident sources

OpsGenie is now a first-class incident source, alongside PagerDuty and Incident.io from earlier weeks. The integration landed end to end across 6 repos in one week: credentials in anyshift-commons, intake and CRUD in anyshift-backend, the agent path in ai-workbench, lambda and fargate wiring in agent-runtimes and agent-gateway, and the UI in anyshift-frontend. Add an OpsGenie credential under the incident integrations page and Annie starts investigating OpsGenie alerts the same way it does PagerDuty ones.

Sentry and Notion links auto-resolve in chat

Paste a Sentry issue URL or a Notion page URL into Annie chat and it now becomes part of the investigation context, the same way a paragraph of pasted evidence would. Annie pulls the Sentry issue through the queryable source that shipped earlier this month, and pulls the Notion page through the lambda MCP that landed last week. No separate paste-and-explain step.

Pull request suggestions from Annie, per hypothesis

Each Annie hypothesis now carries its own "ask Annie for a fix" CTA, and the response comes back framed as a pull request suggestion: here is what I think happened, and here is the change that would address it. The backend exposes a per-project propose-fix endpoint, the frontend renders the CTA on each hypothesis card, and a worker persists the proposal artefact so the same suggestion is reproducible. A hypothesis no longer stops at "here is what I think happened"; it gives the operator a candidate PR to evaluate.

Also this week

  • Chat history filters by source and Slack channel: the chats table now filters by source and by Slack channel, so finding the conversation that paged you on a busy day stops being a scroll problem.

FAQ

What's new at Anyshift the week of May 18, 2026?

Annie gained report personas (concise vs elaborate) and per-conversation effort levels. OpsGenie joins as a new alert source. Pasting a Sentry or Notion link into chat now anchors the investigation. Each hypothesis carries a pull request suggestion from Annie.

How do proactive report personas work?

Open the proactive reports page and pick a persona for the report, for example concise vs elaborate. Annie carries that tone through the report pipeline, so a project pointed at a dev team can run a different style than one pointed at SRE or juniors without two pipelines. The same persona menu also lives in the chat header so a single conversation can override the report default.

What is the effort setting?

Effort controls Annie's performance vs speed trade-off on a per-investigation basis. Set it at the project level under settings, or override it on a single chat from the chatbar. A higher effort means Annie spends more reasoning budget per hypothesis.

What does the OpsGenie integration unlock?

Anyshift now ingests OpsGenie alerts the same way it ingests PagerDuty and Incident.io. Annie investigates with the OpsGenie context alongside your other incident sources, end to end across backend, ai-workbench, agent-gateway, agent-runtimes, and the frontend.

What are pull request suggestions from Annie?

Each hypothesis Annie produces now carries a per-hypothesis CTA that asks Annie for a candidate fix as a pull request suggestion. The backend exposes a per-project propose-fix endpoint, the frontend renders a button on each hypothesis card, and a worker persists the proposal artefact so the suggestion is reproducible.

That's the week. Try it and tell us what it unlocks for your team. We read every reply at app.anyshift.io.