# Keygraph

[Product introduction] Keygraph "Agentic static analysis" combined with "automatic penetration test" can not only scan the code, but also automatically attack the running application. Its core principle is “no PoC without reporting”, and the system only outputs high-confidence vulnerabilities that are verified by real attacks and have reproducible proof-of-concept. In addition to the deep integration mode, it also offers a “black box test” mode where users simply provide a URL to allow the AI to navigate and execute attacks autonomously like a human penetration tester. [Product Features] Intelligent Code Attribute Diagram Analysis: Converts code into CpG structure, and uses LLM to evaluate whether specific desensitization logic can truly defend against specific attacks, significantly reducing false positives. Thirteen professional agents: Built-in 13 dedicated agents, covering five stages of pre-scouting, automated attack surface mapping, vulnerability analysis, and empirical attacks. Parallel exploit system: For OWASP core categories such as SQL injection, XSS, SSRF and Authz, the 5 group parallel agent tries to inject the real payload. Multi-Dimensional Permission Testing: Supports configuring multiple sets of credentials to automatically test vertical and horizontal privilege escalation across organizations, roles (IDOR

[Product introduction] Keygraph "Agentic static analysis" combined with "automatic penetration test" can not only scan the code, but also automatically attack the running application. Its core principle is “no PoC without reporting”, and the system only outputs high-confidence vulnerabilities that are verified by real attacks and have reproducible proof-of-concept. In addition to the deep integration mode, it also offers a “black box test” mode where users simply provide a URL to allow the AI to navigate and execute attacks autonomously like a human penetration tester. [Product Features] Intelligent Code Attribute Diagram Analysis: Converts code into CpG structure, and uses LLM to evaluate whether specific desensitization logic can truly defend against specific attacks, significantly reducing false positives. Thirteen professional agents: Built-in 13 dedicated agents, covering five stages of pre-scouting, automated attack surface mapping, vulnerability analysis, and empirical attacks. Parallel exploit system: For OWASP core categories such as SQL injection, XSS, SSRF and Authz, the 5 group parallel agent tries to inject the real payload. Multi-Dimensional Permission Testing: Supports configuring multiple sets of credentials to automatically test vertical and horizontal privilege escalation across organizations, roles (IDOR

Canonical: https://uniquebase.ai/products/keygraph
Stable identifier: UB-D7866
Snapshot captured: 2026-11-04T00:48:01.362Z
English updated: 2026-11-04T01:06:52.065Z

## Connected profiles

- [Keygraph](/companies/keygraph)

## Sources

- No source links are available for this profile.