Anyshift retweeted
SOTA level: the PR is still open in the MCP spec repo 😅
We shipped experimental Skills over MCP in @anyshiftio Graph MCP anyway.
Codex found the remote skill, read its guidance and used it on a blind, authenticated query.
Happy Friday 🔥
Versioned graph of the infrastructure, not telemetry. That's the architectural split Thomas Prommer flagged when adding Anyshift to his Best AI SRE Tools 2026 guide. ✨
wetheflywheel.com/en/guides/…
2
569
We shipped the Notion integration in @anyshift_io. Annie reads Notion directly now.
The hard part of any knowledge base is search, especially when you have the wrong service name in your head. 🧵👇
1
2
286
Action items came back tagged: two done, two still open, page anchor for each. The one we should actually do this week, alerting on `NumberOfMessagesDeleted` dropping to zero for >1h, is a Linear ticket now.
1
37
Setup is one paragraph: one @NotionHQ internal-integration token gives @anyshift_io's Annie the full Notion workspace, not a hand-rolled REST subset. When Notion ships, Annie has it.
Full write-up: anyshift.io/blog/annie-reads…
35
Two weeks ago our graph-injector queue backed up. The April post-mortem was sitting in our Notion workspace. The on-call still spent ten minutes finding it: four candidates with "graph" or "connector" in the title, a wrong turn through the architecture diagram.
1
21
Five tabs to triage one Sentry error: email, Sentry UI, GitHub for the offending release, deploy dashboard, Slack to ping whoever shipped it.
We just shipped the Sentry integration in @anyshift_io. Annie reads Sentry directly now.
🧵👇
1
3
72
Follow-up: "dig into the agent-gateway MarshalerError." Annie pins it to v0.19.0, persisted through v0.21.22, never fixed.
Then a weekly digest by service and release. graph-connector v0.33.6 critical at the top, 49 new issues this week.
1
16
Setup is a Sentry user-auth token reverse-proxied to @getsentry's official MCP server at mcp.sentry.dev. Every tool Sentry ships, Annie has.
Full write-up: anyshift.io/blog/annie-reads…
1
21
Most teams don't have a real map of their Kubernetes cluster. The wiki diagram is stale, the dashboards only show what someone instrumented, and the first time you find a fragile dependency is the postmortem.
🧵👇
1
1
38
Follow-up: "Which service has the biggest blast radius if it goes down?"
Annie traces the downstream path through frontend and checkoutservice to the user. Same graph, now framed as operational risk surface.
1
13
One click on Save as Recurring Report turns it into a monitor that re-runs on a schedule. Next deploy that quietly removes a probe gets caught before the next incident.
Try Annie: anyshift.io?utm_source=x&utm…
23


