LaunchSafe is a security company that builds AI agents for automated penetration testing. Its platform tests web applications, iOS and Android apps, APIs, and LLM applications and agents, examining each surface for vulnerabilities and attack paths that cross between systems. Tests cover areas such as injection, access control, authentication, server-side request forgery, business logic, and prompt injection.

Teams connect a code repository from GitHub, GitLab, or Bitbucket, then launch agents sized to their codebase. LaunchSafe aims to deliver replayable exploit proofs alongside fixes that can be submitted as pull or merge requests. Findings and fixes can be reported through tools including Slack, Linear, Jira, and PagerDuty.

The platform is designed with customer-defined scope, allow-lists, rate limits, and blast-radius limits; intrusive or state-changing actions require approval, and agent activity is logged. LaunchSafe says its products can run in a customer’s VPC or a nominated region and that customer data is not used to train models. The company notes that security testing is probabilistic, so findings require human review.