One of Bumblebee’s defining features is that it never executes package managers or installation scripts. Instead, it reads metadata already present on the machine—such as dependency lockfiles or installed package records.
This design provides a critical safety benefit during incident response. Many malicious packages hide harmful behavior in installation hooks or post‑install scripts. If a security tool triggers those processes while investigating a system, it could accidentally activate the malware it is trying to detect.
Bumblebee avoids that risk by:
The result is a safe, passive inspection model that surfaces exposure without modifying the environment being analyzed.
Bumblebee gathers an inventory of developer‑machine components that frequently appear in supply‑chain attacks. These include:
The scanner reads package metadata from common language ecosystems, including:
By parsing lockfiles and package metadata, Bumblebee can determine which packages and versions are present on a machine without running the package manager itself.
Developer tools and code editors often run extensions that access source code, tokens, or developer credentials. Bumblebee inventories these editor extensions and manifests to identify risky or compromised plugins.
Security researchers increasingly treat browser extensions as part of the developer supply chain. Bumblebee can inventory browser extensions present on a system to identify those linked to advisories or malicious activity.
A newer attack surface comes from AI development tooling. Bumblebee scans configuration files for AI agents using the Model Context Protocol (MCP) and related tools. Examples include files such as mcp.json and other AI tool configuration formats.
These configurations can reference external tools or services that may introduce supply‑chain risk if compromised.
Bumblebee supports multiple scan profiles so organizations can tailor scans to routine monitoring or incident response scenarios.
A lightweight inventory scan of standard developer‑machine locations. Teams can run this periodically using device‑management or fleet‑management systems.
A targeted scan focused on development directories or specific repositories. This helps teams inspect the packages used in active projects.
A broader investigation mode typically used during an active security incident. It searches wider filesystem locations to find all possible exposures.
These profiles allow teams to scale from routine visibility to full incident‑response sweeps without changing tools.
Bumblebee’s detections rely on exposure catalogs—lists of known risky packages, versions, extensions, or configurations. When the scanner inventories a machine, it compares discovered components against this catalog to find matches.
Each detection is traceable, showing:
This approach helps security teams quickly answer a critical incident‑response question: Which developer machines are exposed to this vulnerability or malicious package right now?
Supply‑chain attacks targeting open‑source ecosystems have grown dramatically. Security research has identified over 1.23 million malicious open‑source packages, with more than 454,000 newly discovered in 2025 alone. Other reports found a 73% increase in detected malicious packages in 2025 compared with the previous year.
At the same time, developer machines now run far more components than traditional security tools track—package managers, editor extensions, browser plugins, and AI agent integrations. Many organizations have limited visibility into this local environment.
Bumblebee addresses that gap by providing fast, safe inventory scanning of developer endpoints. Rather than replacing runtime security or repository scanning, it adds a missing layer of visibility that helps teams quickly detect exposure during supply‑chain incidents.
Perplexity’s Bumblebee brings a practical approach to a difficult security problem: identifying risky components on developer machines without triggering malicious behavior.
By combining a read‑only scanning model, coverage of modern developer tooling (packages, extensions, and AI configs), and catalog‑based detection workflows, it gives security teams a way to rapidly assess supply‑chain exposure across developer endpoints when incidents occur.
As software ecosystems continue to grow—and attackers increasingly target developer tooling—tools that safely inventory the developer environment are becoming an essential part of modern application security.
Comments
0 comments