Technical deep dives into Anyshift's engineering decisions, architecture, and lessons learned.
26 Articles in this series
Break our demo infrastructure on purpose and watch the root cause surface
Sever the link between a service and its database in our new Playground, and a change event lands. Seconds later the root cause comes back, traced against the topology graph instead of a wall of logs. A hands-on way to see what change-first root-cause analysis does, no signup.
Anyshift meets CrowdStrike: safer threat response with production context
CrowdStrike Falcon helps security teams decide what to do with suspicious domains, IPs, and files. Anyshift shows which services, owners, dependencies, and recent deploys are behind the signal before analysts detect, block, or escalate it.
Anyshift meets Confluent: production context before schema versions go live
Confluent is where teams manage Kafka topics, schemas, and data contracts. Anyshift adds the production truth behind a schema change: which producers, consumers, services, owners, consumer groups, and monitors depend on the stream before the version is registered.
AI Context for Prod, Optimized by AIs in Prod
Annie (our AI SRE agent) had institutional memory from ACE, the agentic context-engineering loop that curates cheatsheets from past runs. It worked, but clients kept catching her trusting stale entries or missing answers buried in her own bloated context. So we added five things on top: (1) a fixed set of memory items always presented to the agent, (2) per-query retrieval over the rest of the memory store, (3) an agent-optimized index of that store, (4) the ability for the agent to query the store mid-run, and (5) tried-and-true memory freshness mechanisms. Production context, now optimized by the AI using it. Here's the reasoning and what a few weeks in production say.
Anyshift meets Coralogix: turning telemetry into reviewed production handoffs
Coralogix is where SREs investigate telemetry. Anyshift adds the production graph around a signal: affected service, owner, recent deploy, dependency evidence, and skip reasons, then writes the reviewed handoff into a Coralogix Custom Dashboard.
How we turned on-call judgment into skills an AI agent can load
An AI agent in the incident channel can run kubectl and read a dashboard. What it can't do is judge whether the last deploy is the suspect or a red herring. We open-sourced the SRE skills that encode that judgment, runnable offline against fixtures with no credentials.
Anyshift meets MongoDB Atlas: production-aware alert settings
MongoDB Atlas can alert when a cluster nears its connection limit. Anyshift adds the pre-enable review: affected services, owners, monitors, recent changes, and non-production exclusions before paging starts.
Anyshift meets Snowflake: production context before agents act
Snowflake is where teams govern data, workloads, and AI workflows. Anyshift adds the live production graph those workflows need before they apply a fix, rerun a task, refresh a dynamic table, or trigger an agentic workflow.
Anyshift meets Databricks: checking production impact before a data pipeline rerun
Databricks gives teams the governed data and AI surface. Anyshift adds the live production context a Databricks workflow needs before it patches a data pipeline, reruns a backfill, or calls an agent tool.
Anyshift meets GitLab: production impact before merge
GitLab shows reviewers the diff, pipelines, and approvals. Anyshift adds the missing production layer: which live services use the changed code, who owns them, what can be skipped, and who should review before merge.
Anyshift meets Okta: production reachability for access changes
Okta is where teams manage identity, access, and policy. Anyshift adds production reachability enrichment around an access change: which services, cloud roles, Kubernetes workloads, monitors, and owners sit behind the group before Okta performs the assignment.
Anyshift meets Elastic: debug with PR context already attached
Elastic gives teams the place to search, triage, and open Cases (Kibana investigation tickets) when an incident starts. For a PR that changes a shared authentication module, Anyshift adds what Elastic cannot infer from the PR alone: which production services depend on it, who owns them, Identity hints, and evidence. So when a human or agent starts debugging, the context is already attached.
Anyshift meets Splunk: reduce maintenance alert fatigue
Planned maintenance often creates alert noise. Anyshift finds the Splunk alerts affected by a change, pauses only those saved searches, and turns them back on when the window ends. Teams keep real alerts visible while expected noise stays out of the way.
Anyshift meets Dynatrace: graph context for every deploy
A deployment event should carry the service, owner, and monitored entity it actually changed. Anyshift adds that production context to Dynatrace so on-call teams do not rebuild it from CI and infrastructure tabs.
Anyshift meets New Relic: change impact on every affected entity
Teams investigate incidents in New Relic, but deploy context often lands only on the service that changed. Anyshift maps the real production impact, so every affected New Relic entity gets the deployment context.
Anyshift meets Sentry: releases that follow impact
Sentry is where teams debug regressions. Anyshift makes sure the release context reaches every affected project, including downstream services that did not deploy.
Anyshift meets acli: PR impact, routed into Jira
A shared-code PR should not surprise downstream teams after merge. Anyshift finds the running services and owners affected by the change, then routes the advisory work into Jira before the review is over.
Anyshift meets pup: turning intent into audited Datadog runbooks
Datadog pup can mute monitors during maintenance, but teams still have to know which downstream services will be noisy. Anyshift CLI maps the affected services from production context, then prepares the Datadog downtime runbook with an audit trail.
Anyshift meets gcx: cloning Grafana observability with audited runbooks
Grafana shows the service you instrumented, but downstream services often miss the same dashboards and SLOs. Anyshift maps the dependency graph, finds the coverage gaps, and prepares the Grafana resources for gcx to apply after review.
Annie reads Linear now
Forty minutes paging Linear to confirm a returning customer report was the same bug we'd half-shipped a fix for in February. The Linear integration went GA May 13, and Annie pulled both tickets, the linked PR, and the stalled action in twenty-three seconds.
Annie searches Notion now
Ten minutes to find a post-mortem already sitting in Notion. The Notion integration shipped May 12, and Annie picked the same page in eighteen seconds, root cause and open action items tagged.
How we now know which commit broke each Sentry error
Five Sentry tickets in one worker turned out to be one bug. The most-recent error came from the very PR that had wired Sentry forwarding in. How a stack frame now leads to the offending commit, the deploy behind it, and the team that owns the failing path.
Report Templates: pre-built investigations, one click
Every Monday, the pod-stability review gets rebuilt from scratch. Same dashboards, same correlation work, same write-up. Two hours, gone. Report Templates turn the recurring investigations platform and SRE teams run by hand into one click.
Annie CLI
136 CloudWatch alarms vanish overnight. Annie cross-references Slack, the audit trail, and your infra graph in one query. Now it runs in your terminal.
Agentic Context Engineering in Production: How AI Agents Build Institutional Expertise
AI agents start every run from scratch. ACE (Agentic Context Engineering) gives them institutional memory that evolves through use, cutting root cause analysis time by 30%.
Building a Temporal Infrastructure Knowledge Graph: A Year of Working with Neo4j at Scale
How Anyshift chose Neo4j for building a temporal infrastructure knowledge graph and lessons learned over a year of production use.