Eyry is a security engineering shop that runs the full application security loop. We build production AI security systems, break them with human-led pentesting, harden findings into fixes and guardrails, and watch your attack surface around the clock.
Enter at any stage — ship an AI build, commission a pentest, fix what a test found, or put your surface on continuous watch. See the pipeline.
Security isn't one engagement — it's a loop. Enter anywhere; every stage feeds the next.
Production AI security systems: agent pipelines, eval harnesses, detection and response automation. Fixed scope, source included.
Human-led penetration testing of web, APIs, networks, and LLM systems. Real operators, real exploitation — not a checkbox scan.
Turn findings into fixes: patched code, guardrails, detectors, and automation — so the same bug class stops coming back.
Continuous recon and attack-surface monitoring. See your attack surface the moment it changes — not at the next scheduled test.
A pentest that finds nothing to fix is rare. A build that never gets tested is risky. The loop is the point.
A product you can run yourself or have us host, and a team you can hire.
Open-source tools for recon and offensive security, free to self-host. Or let Eyry Pro run the whole pipeline for you: always-on monitoring, agent runs, alerting, and teams.
Production AI security systems, designed and built for your team — plus pentesting and fractional engineering. The people who build the tools also do the work.
Composable tools for bug bounty hunters, red teamers, and pentesters. Each works on its own; together they watch the surface, dispatch agents, and strike. MIT licensed.
Fast, multi-threaded HTTP prober. The forward scout.
View repo →Producer of new hosts from pluggable feeds (certstream first). The lookout.
View repo →Redis priority queue and work distributor (hot/warm/cold/DLQ).
View repo →Multi-turn agent runtime: compaction, tools, Docker sandbox.
View repo →AI security scanner and reviewer built on Pinnace. The strike.
View repo →Agent control plane: scheduler, events, IRC-style chat, orchestration.
View repo →Eyry Pro is the hosted platform: managed 24/7 attack surface monitoring, agent runs in managed sandboxes, alerting, history, and team features. No infrastructure to stand up.
The open source stays free and uncrippled. Pro sells time, scale, and collaboration.
Early access, launch pricing, and a say in what we build.
Hands-on security engineering from the team building the tools. Each service maps to a stage of the pipeline — enter anywhere, and stages feed into each other.
We design, build, and ship production AI systems for your security team: agent pipelines, eval harnesses, detection and response automation. Fixed scope, source included.
Adversarial assessment of your AI systems and agents: prompt injection, jailbreaks, data leakage, OWASP LLM Top 10. Findings report and 30-day retest included.
Human-led testing of web apps, APIs, networks, and LLM systems. Real operators doing real exploitation — our own tools handle the mechanical work so humans spend their time on judgment calls.
A part-time security engineer embedded in your team: hardening sprints, custom tool builds, ongoing testing, and continuous monitoring. The whole loop, on retainer.
Transparent starting prices. Engagements scale with scope: builds $5k–$20k, AI assessments from $10k, pentests $3k–$12k, fractional $5k–$10k/mo. Every engagement includes a 30-day retest.
Production AI systems, designed & shipped
Adversarial review of your AI systems
Human-led security testing
Your security engineer, on retainer
Fixed price, fixed timeline. No hourly black box. Start at any pipeline stage.
30 minutes. We map your problem, your stack, and what done looks like. Free.
Scope, timeline, and price in writing within 48 hours.
Pentest, build, or hardening sprint — with check-ins so you see it come together.
Findings, fixes, source code, and docs. 30-day retest included.
Whether you need a build, a break, or backup for your own team — talk to the people who'll actually do the work.
Prefer email? hello@eyry.io
Six MIT-licensed tools for recon and offensive security — the same stack we run on real engagements. Try it now, free, today.
This is the team you'd be hiring.
Straight answers on AI security engineering, pentesting, and continuous monitoring.
It's building security into AI systems from the start: agent pipelines with guardrails, eval harnesses that catch regressions, and detection automation that runs in production. We design and ship these systems for your team — fixed scope, source included, from $5,000.
Our human-led pentests start at $3,000 for a focused scope, with most engagements landing between $3k–$12k depending on targets. Every engagement includes a 3–10 day turnaround, a detailed findings report, and a 30-day retest.
Point-in-time testing misses everything that ships after the test. Continuous monitoring watches your scope around the clock — new hosts from certificate transparency, what changed, what matters — so you see your attack surface the moment it changes. Eyry Pro (now on the waitlist) runs it for you, or self-host the open-source suite.
No. We pentest web apps, APIs, and networks too, and build custom security tooling for any stack. The pipeline is the same — build, break, harden, watch — and you can enter at any stage.
The six tools — Vedette, Foretop, Purser, Pinnace, Aplomado, Quarterdeck — are MIT-licensed and free to self-host, uncrippled. Eyry Pro is the hosted version: managed 24/7 monitoring, agent runs, alerting, and team features. Pro sells time and scale, not features.