Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
Python agentic-aiai-securityalert-triageblue-teamcybersecuritydetection-engineeringincident-responselanggraphllm-agentsmcpmitre-attackmodel-context-protocolpurple-teamsecurity-operationsself-hostedsiemsoarsocthreat-detectionthreat-intelligence
---
title: "Example Customer Inc."
industry: "FinTech"
region: "EU (Ireland)"
logo: "/customers/example/logo.svg"
challenge: "Reduce a multi-thousand-alert daily backlog without losing forensic depth."
result_numbers:
- { metric: "MTTR", before: "4h 12m", after: "12m" }
- { metric: "Alerts triaged / day", before: "5,000", after: "320" }
- { metric: "Auto-resolved", before: "0%", after: "78%" }
quote: "We replaced our manual triage queue with an auditable agent loop and our analysts finally stopped chasing duplicates. The investigation ledger settled three SOC 2 questions in one review."
quote_role: "Head of SOC"
quote_company: "Example Customer Inc."
order: 1000
draft: true
related_features:
- { label: "Investigation ledger", href: "/why-open-source#investigation-ledger" }
- { label: "Alert reduction benchmark", href: "/benchmark" }
- { label: "MITRE ATT&CK coverage", href: "/#mitre" }
---
> **Example / Template — replace before publishing.** This file exists so the
> `/customers/[slug]` route renders end-to-end during local development. Every
> number, quote, role, and company name on this page is fabricated. Copy this
> file, swap the frontmatter, rewrite the body, and remove `draft: true` to
> publish a real case study.
## The setup
Example Customer Inc. operates a regulated payments stack across two AWS
regions and a Microsoft 365 tenant. Before AiSOC, the SOC was running a
queue of roughly 5,000 alerts per day across two SIEM tools and a managed
EDR — most of them duplicates of the same underlying entity behaviour.
## What changed with AiSOC
The team enabled the connectors for AWS CloudTrail, M365 audit, and the
EDR. Risk-Based Alerting collapsed entity-level activity into a single
queue; the autonomous triage agent took the first pass on every promoted
alert; and the investigation ledger captured the prompt, tool calls, and
evidence for every decision.
Within a sprint:
- The daily review queue dropped from ~5,000 to ~320 entity-level alerts.
- Mean time to resolution fell from over four hours to twelve minutes for
the bulk of routine cases (the L1 / L2 substitute path).
- 78% of triaged alerts auto-resolved with confidence-gated playbooks, with
every action signed and replayable in the ledger.
## What the auditor saw
The SOC 2 Type II auditor opened the investigation ledger directly and
walked five sampled cases end-to-end: the prompts the agent issued, the
evidence rows it cited, and the actions it executed. No vendor summary.
That review closed three controls (CC7.2, CC7.3, CC7.4) in a single pass.
## What's next
Example Customer Inc. is rolling AiSOC out to its second business unit
and contributing two detection rules upstream — both expected to land in
the next community release.