# What shipped at Anyshift

Week of May 4, 2026

Datadog gains 50+ Bits AI capabilities; PagerDuty + Incident.io + Sentry join as sources; k8s-agent v0.3.2 brings on-demand graph reconciliation.

Author: Anyshift

Published: 2026-05-11

Source: https://anyshift.io/blog/anyshift-shipped-week-of-2026-05-04

Here's what we shipped to customers during the week of May 4, 2026: Datadog gains 50+ new capabilities via Bits AI, PagerDuty / Incident.io / Sentry join as new sources, anyshift-k8s-agent v0.3.2 brings on-demand graph reconciliation, the Annie Automation editor adds a simulator and group fallbacks, reports auto-inject project knowledge, and the Anyshift MCP server now ships Cypher graph-query expertise to any IDE agent. Across 10 product repos, the team merged 55 PRs and cut 10 backend releases (v0.83.23 through v0.84.1) in the [Anyshift Product Updates](/blog/series/anyshift-product-updates) series.

## Integrations Annie can query

### Datadog: 50+ new capabilities via Bits AI

[Annie](/product) now investigates with [Datadog](/integrations/datadog) Bits AI capabilities (Watchdog stories, error tracking, AI trace summaries, dashboard-deep analysis, and 50+ more) without us reimplementing any of them. We route through Datadog's official MCP server, gated behind a `DatadogOfficialMCP` feature flag with a server-side write blocklist for safety. The warm-up prompts in `anyshift-backend` now harvest the Bits AI catalog automatically.

### PagerDuty and Incident.io ingestion

[PagerDuty](/integrations/pagerduty) and [Incident.io](/integrations/incident-io) are now first-class incident sources. Their webhooks ingest silently, join with Annie Automation routing rules, and show up in a dedicated admin/router PagerDuty tab. Combined with the new GroupFallback model, you can route incidents to different Annie behaviors per source group.

### Sentry as a queryable source

[Sentry](/integrations/sentry) joins the MCP-accessible lineup via a stdio-based lambda MCP. Annie can now reach Sentry projects through Sentry's own MCP server when investigating an RCA.

## Kubernetes graph reconciliation in anyshift-k8s-agent v0.3.2

Upgrade your installed [anyshift-k8s-agent](/integrations/kubernetes) to v0.3.2 (tagged 2026-05-05) to unlock on-demand graph reconciliation. When your cluster state drifts (new nodes, freshly applied manifests, recent rollouts), our team can trigger an immediate reconciliation against your live cluster so Annie investigates from the current state, not the last scheduled sync. The agent exposes a new gRPC endpoint for the refresh; the audit trail and trigger stay on the Anyshift side.

## Annie Automation: simulate before you save

[Annie Automation](https://docs.anyshift.io/pages/product/customization/instructions) gained 3 new primitives:

- **Test-a-message simulator**: dry-run a rule against a sample event before saving. No more push-to-find-out that a routing decision was wrong.
- **Per-group fallbacks**: configure different default behaviors per `(sender, channel)` group, layered under a single rule.
- **Rule conflict evaluator**: when two rules could match the same event, the editor flags the overlap before you save.

## Reports auto-include project knowledge

Custom and proactive reports now automatically inject project-level knowledge, so a report scoped to a specific service or namespace stays anchored to that context. The previous workflow required restating the project context in each report; that's gone. This pairs well with last week's [Slack reports rollout](/blog/anyshift-shipped-week-of-2026-04-27): the report you fan out to a channel is now correctly scoped without manual prompt engineering.

## Cypher graph-query expertise, now in any IDE agent

Anyshift's [remote MCP server](https://docs.anyshift.io/pages/product/integration/remote_mcp) (`@anyshift/anyshift-mcp-server` v0.10.2) now ships the full Cypher graph-query prompt directly via MCP tool descriptions and instructions. Any agent that speaks MCP (Claude, Codex, Cursor, your own) inherits the same graph-query fluency Annie uses, with zero per-agent prompt engineering. Pull the latest npm release to upgrade.

## FAQ

### What's new at Anyshift the week of May 4, 2026?

Datadog gains 50+ new capabilities via Bits AI (Watchdog stories, error tracking, AI trace summaries, dashboard-deep analysis). PagerDuty and Incident.io join as new incident sources. Sentry is now a queryable source via MCP. `anyshift-k8s-agent` v0.3.2 brings on-demand graph reconciliation. Annie Automation got a message simulator with per-group fallbacks, and custom and proactive reports auto-include project knowledge.

### How did the Datadog integration improve this week?

Annie now investigates with Datadog Bits AI capabilities: Watchdog stories, error tracking, AI trace summaries, dashboard-deep analysis, and 50+ more. We route through Datadog's official MCP server, gated behind a `DatadogOfficialMCP` feature flag with a server-side write blocklist for safety.

### Can Annie ingest PagerDuty and Incident.io alerts?

Yes. PagerDuty and Incident.io are now first-class incident sources for Annie. Their webhooks land silently, get joined with custom-instruction routing rules, and surface in the admin/router PagerDuty incidents tab.

### What does anyshift-k8s-agent v0.3.2 unlock?

Upgrading to `anyshift-k8s-agent` v0.3.2 unlocks on-demand Kubernetes graph reconciliation. When your cluster state drifts (new nodes, fresh manifests, recent rollouts), our team can trigger an immediate reconciliation against your live cluster so Annie investigates from the current state, not the last scheduled sync.

### How do Annie's reports get smarter this week?

Custom and proactive reports now auto-inject project-level knowledge, so a report scoped to a specific service or namespace stays anchored to that context. No manual prompt engineering per report.

### What changed in Annie Automation this week?

Annie Automation got a test-a-message simulator, per-(sender, channel) group fallbacks, and a rule conflict evaluator. You can now dry-run a rule before saving it and configure different fallback behaviors per source group.

---

That's the week. Try it and tell us what it unlocks for your team. We read every reply at [app.anyshift.io](https://app.anyshift.io).

## What's new at Anyshift the week of May 4, 2026?

Datadog gains 50+ new capabilities via Bits AI (Watchdog stories, error tracking, AI trace summaries, dashboard-deep analysis). PagerDuty and Incident.io join as new incident sources. Sentry is now a queryable source via MCP. anyshift-k8s-agent v0.3.2 brings on-demand graph reconciliation. Annie Automation got a message simulator with per-group fallbacks, and custom and proactive reports auto-include project knowledge.

## How did the Datadog integration improve this week?

Annie now investigates with Datadog Bits AI capabilities: Watchdog stories, error tracking, AI trace summaries, dashboard-deep analysis, and 50+ more. We route through Datadog's official MCP server, gated behind a DatadogOfficialMCP feature flag with a server-side write blocklist for safety.

## Can Annie ingest PagerDuty and Incident.io alerts?

Yes. PagerDuty and Incident.io are now first-class incident sources for Annie. Their webhooks land silently, get joined with custom-instruction routing rules, and surface in the admin/router PagerDuty incidents tab.

## What does anyshift-k8s-agent v0.3.2 unlock?

Upgrading to anyshift-k8s-agent v0.3.2 unlocks on-demand Kubernetes graph reconciliation. When your cluster state drifts (new nodes, fresh manifests, recent rollouts), our team can trigger an immediate reconciliation against your live cluster so Annie investigates from the current state, not the last scheduled sync.

## How do Annie's reports get smarter this week?

Custom and proactive reports now auto-inject project-level knowledge, so a report scoped to a specific service or namespace stays anchored to that context. No manual prompt engineering per report.

## What changed in Annie Automation this week?

Annie Automation got a test-a-message simulator, per-(sender, channel) group fallbacks, and a rule conflict evaluator. You can now dry-run a rule before saving it and configure different fallback behaviors per source group.

