Anyshift

Integrations cover AWS, GCP, Azure, Cloudflare, GitHub, GitLab, Kubernetes, Datadog, Grafana, Splunk, PagerDuty, incident.io, Sentry, Jira, Confluence, Notion and Linear. Annie posts RCA results as PagerDuty notes and incident.io comments and opens pull requests through a separate agentic GitHub app. The integrations number 25+, and the graph also reads Terraform code and Terraform state.

SourceAnyshift, docs.anyshift.io Root Cause Analysis, Propose Fix and llms.txt indexread 2026-10-05

Annie runs its own investigation across the graph, logs and traces and returns a timeline, root cause and remediation steps, coordinating specialized agents along the way. Customers tune it with instructions, personas and effort modes, and automation rules route Slack messages with fallbacks for each sender and a fixed default. There is no builder, branching or multi agent flow that customers control.

SourceAnyshift, docs.anyshift.io Root Cause Analysis and llms.txt indexread 2026-10-05

Annie and other agents query one versioned knowledge graph of infrastructure, applications, identity and team tools, with timestamped state changes and entity resolution that joins an EC2 instance, a Kubernetes pod, a GitHub repo, a Jira ticket and a Confluence page into single nodes. The graph links Terraform code, Terraform state and live infrastructure, and change history shows what changed and when before an incident.

SourceAnyshift, anyshift.io and docs.anyshift.io llms.txt indexread 2026-10-05

Annie's only write action, a fix pull request, runs when a person clicks Propose Fix on a hypothesis or answer. Write access is isolated in a separate agentic app while investigation stays read only, and nothing reaches production until a person merges the pull request.

Sourcedocs.anyshift.io/pages/product/propose_fixread 2026-09-29

Role based access control, multi factor authentication and single sign on are supported, with roles and permissions set out in their own documentation. Anyshift holds AICPA SOC 2 Type II, covering security, availability, processing integrity, confidentiality and privacy, and a trust portal is published at trust.anyshift.io. The platform runs on AWS with encrypted storage, data isolation and retention policies, and code is scanned continuously for secrets, vulnerabilities and risky dependencies.

SourceAnyshift, docs.anyshift.io compliance and llms.txt indexread 2026-10-05

Each RCA carries a timeline of events, root cause details with evidence from the customer's systems and affected resources, and fix PRs link back to the conversation, but there is no record of the steps, queries or tools Annie used in the run, and retention is not stated. Saved reports can be reused and scheduled, and change history shows what changed, when, and what came before an incident.

SourceAnyshift, docs.anyshift.io Root Cause Analysis and llms.txt indexread 2026-10-05

After each investigation Annie reflects and curates learnings that inform later investigations, covering infrastructure patterns, normal baselines and error signatures, effective queries and dashboard locations, and which services power which features and who owns them. Users add custom knowledge Annie cannot learn on its own, such as team conventions, vendor specifics and escalation procedures. Anyshift does not say who the learnings apply to, how long they are kept or how to view or delete them.

SourceAnyshift, docs.anyshift.io Memory and Agentsread 2026-10-05

The product is SaaS with a 30 minute setup, and there is no self hosting, VPC, region or residency option. The service runs on AWS.

SourceAnyshift, docs.anyshift.io complianceread 2026-10-05

There is one agent, Annie, with product capabilities (RCA, reports, proactive detection, propose fix) and Graph API use cases (alert noise, configuration churn, pull request blast radius). Those are features of one product, not separate agents or templates a customer adopts. Saved reports can be reused across the team.

SourceAnyshift, docs.anyshift.io llms.txt indexread 2026-10-05

RCA starts automatically when an incident is created in PagerDuty or arrives through a webhook from incident.io, and investigations also start on demand from Slack and the Remote MCP, from pull request analysis and from proactive detection. Saved reports can run on a schedule, and automation rules let Annie act on Slack messages by itself.

SourceAnyshift, docs.anyshift.io Root Cause Analysis and llms.txt indexread 2026-10-05

Anyshift lists its model providers among its external providers, and customers cannot choose the model. An effort mode trades speed against depth of reasoning, and personas set tone, depth and assumptions, but neither changes the model.

SourceAnyshift, docs.anyshift.io compliance and llms.txt indexread 2026-10-05

There is a read only Graph API over HTTP with a TypeScript SDK (@anyshift/graph-sdk), a CLI, a Graph MCP server and a published OpenAPI specification, plus Annie Remote MCP for triggering RCA from an IDE. The Graph API serves code, CI pipelines, internal platforms and other AI agents, and the CLI works from the terminal, scripts and CI.

SourceAnyshift, docs.anyshift.io Graph API and llms.txt indexread 2026-10-05

Pull request blast radius and drift detection check the customer's infrastructure changes before they ship, which tests the change, not the agent. There is no way to test Annie's own answers. Drift detection shows the gap between what Terraform declares and what is actually running.

SourceAnyshift, docs.anyshift.io llms.txt indexread 2026-10-05

Annie works through integrations, read only live queries, Slack, pull requests, an API and MCP, and does not control a browser, desktop or computer. The only write access sits in a separate agentic GitHub app that opens fix pull requests.

SourceAnyshift, docs.anyshift.io Root Cause Analysis and Propose Fixread 2026-10-05