E
Home

/ The AI Hub

How the AI works. In four pieces.

Everywhere Security uses an agent fabric to tailor what you see. Four agents do the heavy lifting — the homepage personalizer, the newsletter author, the risk surfacer, and the CVE advisor. Each agent runs on the WIT OS Dynamic LLM Router, so sensitive prompts route to private LLMs and never leave the perimeter.

/ Agent · Personalizer

It tunes the homepage.

When you create a profile, the personalizer agent maps your industry and stack tags to a subset of feeds, threats, and topics. Each visit, it re-ranks the homepage so you see the items that move the needle for your environment first. No noise. No general-audience headlines.

How it runs
  1. 01Profile created → industry + stack saved (browser-only for now)
  2. 02Personalizer maps to industry threats + stack-relevant CVEs
  3. 03Homepage re-renders with your slice of the feed

/ Agent · Newsletter

It writes you a unique newsletter.

No two subscribers get the same digest. The agent reads your profile, watches the feeds you've subscribed to, and drafts an issue tuned to your reading level, your role, and your cadence. Daily, weekly, or monthly. Topic mix you control.

How it runs
  1. 01Cron tick → query feeds you subscribed to (last 24h / 7d / 30d)
  2. 02Filter by your stack and industry · cite every claim
  3. 03Draft with private LLM (sensitive prompts never leave the perimeter)
  4. 04Deliver via email · archive in your dashboard

/ Agent · Risk surfacer

It tells you which CVEs hit you.

Most CVE advisories are noise — high CVSS on a product you don't run is irrelevant. The risk surfacer cross-references every new CVE with your stack tags, KEV status, and EPSS exploit-prediction score. You get a short list of CVEs that actually need action.

How it runs
  1. 01New CVE published in NVD
  2. 02Match against your stack tags + KEV catalog + EPSS score
  3. 03Prioritize: KEV > EPSS > CVSS
  4. 04Surface only the top items in your dashboard + digest

/ Agent · CVE Advisor

It translates CVEs into operator language.

Plain English. What the CVE actually does. What products are affected. What you should patch, what you should monitor, what you can deprioritize. Every claim cited back to the source advisory.

How it runs
  1. 01CVE selected (by you or the risk surfacer)
  2. 02Agent fetches vendor advisory + exploit POCs + KEV note
  3. 03Generates a 200-word operator briefing with action items
  4. 04Cites every claim · output reviewed before delivery

/ Under the hood

Four agents. One router. Private by default.

n8n orchestration
Cron-driven workflows for ingestion, dedup, and delivery.
Dynamic LLM Router
Sensitive prompts to private LLMs (Llama 3.3, Qwen 2.5). Public prompts to OpenAI, Anthropic, Google.
WIT OS MAESTRO fabric
Hierarchical agent orchestration. Audit trail on every decision.
Edge delivery
Cloudflare Workers + Vercel for the public site. Sub-second TTFB worldwide.
Detect.Respond.Automate.Predict.Defend.Operate.Detect.Respond.Automate.Predict.Defend.Operate.Detect.Respond.Automate.Predict.Defend.Operate.Detect.Respond.Automate.Predict.Defend.Operate.
Detect.Respond.Automate.Predict.Defend.Operate.Detect.Respond.Automate.Predict.Defend.Operate.Detect.Respond.Automate.Predict.Defend.Operate.Detect.Respond.Automate.Predict.Defend.Operate.
WITONE — Innovate Securely

Ready to run on WIT OS?

Talk to the team about a managed deployment, a pilot, or a custom agent — we typically respond within an hour.