MindFort
Recursively learning security agents
60 followers
Recursively learning security agents
60 followers
MindFort is the fastest way to deploy security agents. We secure your stack via autonomous agents that learn as they hack. We validate each vulnerability and give you native ways to patch. Built by a team with 15+ years of Cybersecurity experience. YC X25. Used by top startups + public companies.
the "recursively learning" part is what caught my eye — most security tools run a scan and give you a report, and then you're on your own. the idea of an agent that not only finds vulnerabilities but actually generates patches and learns from each cycle is a very different mental model. i do wonder though how it handles false positives at scale — if the agent is autonomously patching things, a wrong call could create more problems than it solves. is there a human approval step before it touches anything in production?
MindFort
@priyanka_garhwal1 We create a PR for you to review on each patch :)
HillClimb is the part that hooked me , most pentest tools find the same surface-level issues every run because they don't learn. I'm curious how MindFort handles false positives at scale, since a PR for every finding gets noisy fast if validation isn't near-perfect.
MindFort
@sailikhith Each finding comes with steps to reproduce, and how the agent was able to exploit the vulnerability! Each agent also has context about your target, so it lowers false positives to almost none.
MindFort
Hey Product Hunt 👋
Brandon here - founding team @ MindFort. I've spent 10+ years in cybersecurity, and have never seen attacks hit companies at this scale. With models getting better, no one is safe. We built a tool to help companies counter attacks before they even appear, autonomously.
MindFort is the fastest way to deploy and secure your web app with autonomous agents that expose, prove, and fix vulnerabilities. Our agents learn and get better the more you run it, just like a real attacker.
Launching today:
AI Pentests — autonomous agentic assessments. Deploy a team of agents to hack your app, each with their own toolkit. Deploy agents in < 15 mins. First results in hours, not weeks.
AI Red Team — always-on adversarial agents that you can deploy on a schedule.
HillClimb — recursive learning. Agents consolidate memory and compound experience across attacks.
Validated Findings — MindFort gives you the exact steps to exploit and chain vulnerabilities.
Automated Remediation — every finding ships as a PR, with Github, Linear, and Jira integrations.
MF-1 — our custom offensive-security LLM. Built to chain exploits, break auth, and reason like an attacker
Defenders have to win every day. Attackers only need to win once.
In the comments all day. AMA.
— Brandon
The recursively learning/HillClimb part especially caught my attention here. Most pentesting tools still feel very repetitive where every run starts from zero again, so agents that actually compound experience across assessments feels like a genuinely interesting direction.
I wanted to immediately try and test it out using the free tier but unfortunately I couldn't. Upon trying to signup using both the email and Google SSO, I was met with "Access blocked, please contact support" I have attached an image to show the same.
Would genuinely love to try this properly in the future once that gets resolved, especially to see how the recursive learning behaves over repeated runs against the same application.