PortSwigger Introduces Burp AT Agentic AI for Automated Penetration Testing

PortSwigger has launched Burp AT, an agentic AI system that allows penetration testers to delegate web security investigation tasks while maintaining direct control over the testing scope, approvals, and final conclusions. The public beta is currently available for Burp Suite Professional users.

Unlike standalone AI assistants that operate based on prompts with limited context, Burp AT integrates seamlessly with Burp Suite’s existing tools and project data.

It can utilize captured traffic, the structure of target sites, previously identified issues, and discovery information within a Burp project to investigate potential vulnerabilities. This integration enables agents to work alongside testers rather than conducting isolated scans.

PortSwigger Burp AT Agentic AI

Burp AT is built around four key components: Burp Suite tools, project-specific context, specialized pentesting skills, and policy-driven controls.

PortSwigger stated that its agents use specialized capabilities developed with PortSwigger Research, allowing them to follow structured security testing methods instead of relying solely on general knowledge models.

The platform also features configurable autonomy. Testers can determine whether an agent should investigate a lead independently, request approval before taking certain actions, or be restricted from accessing specific tools and targets.

Significantly, scope, tool access, and approval rules are enforced at the Burp tooling level rather than being subject to the AI model’s interpretation.

Features and Technical Details

• Agentic AI for pentesting that can formulate hypotheses, use security tools, interpret results, and determine follow-up actions.
• Native Burp Suite integration, enabling agents to operate within established professional web security testing workflows.
• Project-aware analysis using captured HTTP traffic, target structure, discovered endpoints, and existing issue data.
• Purpose-built pentesting skills developed in collaboration with PortSwigger Research to guide agents through structured security testing methodologies.
• Configurable autonomy, allowing testers to dictate what agents can investigate, what requires approval, and what is blocked.
• Burp-enforced boundaries for target scope, tool access, and approval policies, separate from the AI model’s decision-making mechanisms.
• Evidence tracking that maintains records of requests, responses, actions, and findings within the Burp project for validation and reporting.
• JavaScript analysis capabilities, including the reconstruction of endpoints and workflows from large or minified client-side codebases.
• – Human-led operation, with testers retaining responsibility for scope, security judgment, exploit validation, and final conclusions.
• – Availability of the public beta for Burp Suite Professional users, with future plans for team and enterprise workflows.

PortSwigger noted that Burp AT can manage time-consuming tasks that often exceed the limitations of a conventional engagement.

In a closed beta example, a tester requested the system to analyze 66,000 lines of minified JavaScript, reconstruct referenced endpoints and workflows, and identify sensitive or unauthenticated functionalities. This investigation reportedly uncovered a critical vulnerability that might have gone untested for another year.

The company emphasized that Burp AT is not designed to replace the human tester in the process. Instead, it aims to enhance testing coverage, explore unfamiliar application areas, and validate leads that would otherwise contend for limited pentesting time.

Burp AT marks PortSwigger’s initial phase of agentic AI integration. Future developments are expected to introduce team and enterprise operating modes that feature more autonomous testing under established policies, shared visibility, and enhanced audit capabilities.

ALERT: 20+ government sites delivered malware to businesses and citizens. See full attack research to check your own exposure.