The moment that triggered it

2am on a Tuesday.
And I was re-investigating
something we'd already solved.

// Night shift, MSSP SOC, three years ago
incident-log — analyst session #441

A client called. Credential-based anomaly on their VPN, 11pm local time.

// I'd worked this exact pattern three months prior.

Same actor. Same TTP: cached SSO token replay via browser cookie export. We'd documented it. One of our analysts — let's call him Marcus — had written it up completely: actor ID, IOCs, detection logic, recommended hunting path.

// Marcus left for another job the month before.

His notes existed. Somewhere. In a ticket. In an email thread. In his head, when he was here.

I spent four hours re-establishing context that Marcus had already established. I didn't know his notes existed. I didn't know he'd already correlated the VPN log with the AAD sign-in. I didn't know he'd already called the client back with a full brief.

By the time I found the ticket, it was 4am. The client had been sitting with an open incident for five hours.

// We had the answer. It was in someone's memory.

Every SOC has this problem.

CTI feeds give you facts about the world. TLP markings give you provenance. STIX gives you a data model. What nobody gives you is your team's memory — the institutional knowledge of what you've investigated, what you've found, and what you've decided.

01

Analysts leave

Institutional knowledge walks out the door with whoever generated it. The next analyst starts from zero — or spends hours recreating context that already exists.

02

Context lives in tickets

Findings are scattered across SIEM alerts, Jira tickets, email threads, Slack messages, and shared drives. Cross-referencing them requires human memory — or hours of manual search.

03

No single source of truth

Every analyst has a different mental model of the threat landscape. When two analysts investigate the same indicator, they reach different conclusions — or never realize they're working the same case.

AI agents re-investigate what they already solved.

Every AI agent that touches your SOC starts from a blank slate. It doesn't know what your team found last month. It doesn't know which actor family uses which TTP in your vertical. It doesn't carry context between sessions.

Agent session #441
> Investigate: VPN auth anomaly on 2026-05-23
CONTEXT No prior sessions found

This appears to be an initial investigation. Starting from available CTI feeds only.

FINDING

Anomalous VPN authentication from external IP — possible credential-based intrusion. Recommend: block IP, rotate tokens, review AAD sign-in logs.

Agent has no visibility into prior work. Investigation starts from scratch.
Agent session #442 — WITH ThreatRecall
> Investigate: VPN auth anomaly on 2026-05-23
MEMORY Cross-session recall · 2 prior investigations

Resolved: 2026-02-14. Same IP range. Actor TTP-4412. Cached SSO token replay. Client brief delivered by Marcus Chen, 2026-02-15. IOCs: 185.220.x.x, vpn-client-export.exe hash.

SYNTHESIS

This is a repeat of incident INC-2026-02-14. Same actor. Same TTP. Update detection rule with actor TTP-4412 and rotate client SSO tokens. Confidence: HIGH — 3 evidence chains, 2 analysts.

Agent carries institutional memory. Investigation completes in minutes, not hours.

Not another CTI feed.

ThreatRecall isn't here to tell you what's happening in the world. Feeds do that. MITRE ATT&CK does that. OpenCTI does that.

ThreatRecall is here to give your team a persistent, queryable memory layer — so the next time your AI agent (or your analyst) asks "have we seen this before?" they actually get an answer.