In July, we introduced Savant Pathseeker through an early access program built on the belief that the future of offensive security is about AI and humans working the same attack surface together, with agents delivering breadth and cadence and people bringing the judgment to find business logic flaws, chained attacks, and the vulnerabilities that do the most damage.

Since then, early access customers put Pathseeker to work against real external attack surfaces, and their feedback directly shaped what we built next. 

Today, Savant Pathseeker is generally available. Once you have an On-Demand or Continuous plan, you can scope and launch the tests yourself, with no services engagement required. 

While some approaches to AI pentesting still route every test through an internal services team, with scoping calls, scheduling, and a delivery queue that add days or weeks to work that should take hours. Pathseeker puts your team in control: select approved targets, add guidance, launch the test, and review findings as they land in your Bugcrowd Security Inbox.

“Security teams shouldn’t have to choose between testing their most critical assets deeply and testing everything else often. We believe agentic pentesting should move at the speed of the problem it’s solving,” said Braden Russell, Chief Technology Officer at Bugcrowd. “Savant Vista finds assets and then Savant Pathseeker tests them. Human pentesters and security researchers can go deeper. The human + AI approach is a key part of Bugcrowd’s DNA.” 

What’s new at General Availability

Over the past few months, early access customers ran Pathseeker against real attack surfaces while the bar for agentic testing kept rising across the industry. Release cycles are accelerating with AI-assisted development, and regulators and regulations such as the SEC, DORA, and NIS2 increasingly expect continuous, demonstrable validation rather than point-in-time attestations. The GA release of Savant Pathseeker reflects both.

This evolution includes improving our testing engine and skills library, expanding authenticated testing capabilities, and increased transparency for auditability. All of these bring customers better results, deeper testing, and a more intuitive experience. Here’s a breakdown of some of the biggest improvements. 

Scoping straight from Savant Vista

With bulk asset import and target scoping directly from Savant Vista inventory, assets you’ve discovered can move into testing without manual re-entry. Discovery and testing now work as a single pipeline. Savant Vista is also now generally available. 

Expanded login support

Much of an application’s most valuable functionality sits behind authentication. In addition to form-based login and common SSO/SAML identity provider pages, Pathseeker now supports MFA (TOTP), JWT, and API and bearer tokens, so it can test more of what your applications expose to signed-in users. Test credentials are encrypted, scoped to your account, and kept out of logs and reports.

Severity aligned to Bugcrowd VRT

Findings are now rated using the Bugcrowd Vulnerability Rating Taxonomy (VRT) and standard CVSS scores, with a written rationale explaining each rating. That means your team can prioritize without re-triaging every result, and Pathseeker findings use the same severity scale and appear in the same Security Inbox as findings from your bug bounty, VDP, and pentest programs. 

Downloadable reports and activity logs

Every test now produces a downloadable PDF report and a full activity log showing what the agent did and when, giving increased transparency into attack paths. Results can move beyond the platform to auditors, executives, and other stakeholders, with reproducible evidence attached to each finding.

Plain-language test guidance

Keep your team in control of where agents spend their time. Before a test begins, give Pathseeker custom context and plain-English instructions on what to prioritize or avoid—with more formats on the way.

API-triggered testing

Tests can now be launched through the API, so you can kick off Pathseeker from CI/CD pipelines or your own tooling and fit testing into the workflows your teams already use. 

How Savant Pathseeker works

  • Scope: Select approved external assets from Vista or add them manually. Each test starts with proprietary in-depth discovery and recon against your targets to provide comprehensive context during the test planning process. 
  • Plan: The engine builds a test plan based on industry testing standards, shaped by any guidance you provide.
  • Test in parallel: Specialized agents map the target, handle login, and hunt for specific vulnerability types, including applying our DARPA Grand Championship winning API fuzzing technology. 
  • Work in rounds: Agents review confirmed results before deciding what to test next. Interrupted tests can resume.
  • Prove: Findings include evidence that the attack worked. Where proving it could harm the asset, the finding is labeled a strong indicator.
  • Deliver: Findings arrive in your Security Inbox with VRT severity, a downloadable report, and an activity log.

Tests stay within the scope you define, the engine is designed to be non-destructive, and a kill switch stops a test immediately. Pathseeker runs on frontier AI models with a proprietary skills layer built by Bugcrowd’s in-house practitioners, and Bugcrowd does not use customer or researcher data to train or tune those models. You can learn more in the Pathseeker Overview Documentation. 

Built for your external attack surface

Pathseeker is designed for external, internet-facing web applications and APIs. It strengthens your testing program with continuous, evidence-based coverage.

Pathseeker is also built to work alongside, rather than in place of, the human-led pentests that many compliance frameworks require. This is because AI changes the economics of coverage. It can test the long tail of your external estate at a cadence no human team can match, freeing your human testing budget for novel attack chains and business logic. 

Get started today

There are three ways to get started:

  • On-demand: A set of tests to use whenever you need them over a one-year term. A good fit for testing a defined group of applications or filling the gaps between scheduled pentests.
  • Continuous: An annual subscription with a monthly test allotment that refreshes every month, built for teams that want ongoing coverage across a larger external estate.
  • Proof-of-value: A set number of no-cost tests to use within 90 days, so you can evaluate Pathseeker on your own scope before committing. Available alongside a new or renewing Vulnerability Disclosure or Managed Bug Bounty program, and for organizations new to Bugcrowd Pen Test as a Service.

Each test covers up to five related targets and runs for up to 24 hours. If a test fails, or you cancel it within the first 10 minutes, it doesn’t count against your allotment. Pricing is predictable and per-test, and higher volumes lower the per-test cost, with no token meter running in the background.

Explore Savant Pathseeker and launch your first test. Already a Bugcrowd customer? Talk to your account team about adding Pathseeker or starting a Proof of Value. 

To everyone who joined our early access program: thank you. Your time, testing, and candor shaped this release.