MindFort runs autonomous penetration testing for engineering and security teams that ship software faster than manual pentests can keep up with.
Swarms of AI agents attack live applications and APIs in parallel, prove each finding with a working exploit, trace it back to code when a repository is connected, and open a patch as a pull request that is retested after deployment.
At A Glance
| Vendor | MindFort |
|---|---|
| Category | Autonomous AI penetration testing and remediation |
| Deployment | SaaS, run on demand, on a schedule or from CI/CD |
| Agentic AI | Agent swarms, MCP control, MF-1 model |
| Customers | Origin, Bluejay, Nixtla, Sentra, 11x |
| Certifications | SOC 2 Type II |
| Founded | 2025 |
| Market classification | Agentic AI Identity |
What MindFort Does
Traditional penetration tests happen once or twice a year, while code changes every day. MindFort’s answer is autonomous agents that find, validate and patch vulnerabilities continuously. The workflow has three stages, deploy, learn and remediate. Agents are pointed at an environment, build up context about it over time and remember which techniques worked, then hand fixes back to developers through the tools they already use. A finding counts as resolved only after a retest shows the exploit no longer works. The longer-term aim is for discovery and patching to run quietly in the background, with a stated mission of defending against a coming wave of cyberattacks run by AI.
Products
Security Swarms
The core product tests live applications and APIs around the clock. Each finding arrives with proof of exploit and steps to reproduce, so developers are not left chasing scanner noise.
- Black Box testing without source code, and White Box testing that reads source to trace data flows
- Authenticated testing behind SSO and MFA
- Business logic testing, and surface mapping that finds forgotten hosts
Patching And Retesting
Once a weakness is confirmed, agents write the fix and open a pull request. After deployment they rerun the original exploit to confirm the hole is closed. Cloud misconfigurations get the same treatment.
Validator
Teams drowning in findings from scanners, bug bounty programmes and past pentests can feed them into Validator. It tries to reproduce each one and rejects those that cannot be exploited.
Coverage, Reporting And PR Security Reviewer
For teams facing audits, the platform tracks test coverage and assembles pentest reports and compliance evidence, with Analytics showing open findings and OWASP Top 10 exposure. PR Security Reviewer reads pull requests line by line before they merge, and custom tasks such as bug report triage can be written in plain language.
How It Works
| Component | What it does |
|---|---|
| Deployment | A hosted service. Tests run on demand, weekly, monthly or on CI/CD triggers. |
| Agents | Agents can be mentioned in Slack, or reached through the dashboard, the API or MCP, and they accept custom tasks. |
| Agentic Control System | Every change an agent makes to infrastructure is versioned with its before and after state, routed through approval and can be rolled back. |
| Testing approach | Agents chain exploits together and combine white-box access to code with black-box exploitation of the running application. |
| Audit trail | Each remediation records who, what, when and why. Agents propose changes and the customer’s team approves them. |
Identities And Systems Covered
| Applications | Live web applications and APIs, tested with or without source code |
|---|---|
| Authenticated surfaces | Application areas behind SSO and MFA |
| Code | Source repositories on GitHub, used to trace findings and open pull requests |
| Cloud | Configuration remediation on AWS, Azure and GCP |
| Attack surface | Forgotten hosts, which are added to the target inventory once found |
Agentic AI Capabilities
Agent Swarms And MF-1
Many agents work in parallel against one target. MF-1 is a custom agent framework that lets them use tokens efficiently for offensive work, and each agent can write exploit scripts, validate vulnerabilities and open pull requests or tickets on every push.
MCP And API Control
Agents use the Model Context Protocol to discover and call tools, so scanners, code hosts, ticketing and cloud APIs connect without custom glue code. A newly added MCP server is available to agents on their next run.
Context And Memory
Agents take in customer documentation, specifications and accepted risks as context, and carry what worked into later runs.
Integrations
| Area | Integrations |
|---|---|
| Ticketing | Jira, Linear |
| Collaboration | Slack |
Company Background
| Founded | 2025 |
|---|---|
| Location | San Francisco |
| Founders | Brandon Veiseh (Co-Founder and CEO) and Akul Gupta (Co-Founder and CTO) |
| Funding | A reported $3M+ seed round in April 2026, led by Soma Capital with Y Combinator, 468 Capital, CRV, Sandwith Ventures and Blast |
| Certification | SOC 2 Type II |
Note: This overview is compiled from MindFort’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.