Anthropic Releases Claude Security Plugin to Hunt Bugs in Codebase
Artificial intelligence is officially taking over the job of tired senior developers who skim pull requests at 5 PM on a Friday. Anthropic just turned code auditing into an automated team effort right inside the local terminal session.
The new Claude Security plugin operates directly within local Claude Code sessions for paid subscribers. Instead of relying on a single AI model to spot flaws, the tool deploys a multi-agent network where specialized virtual assistants analyze project architecture, construct threat models, and cross-check every single vulnerability finding before adding it to the final report.
Developers can set these digital security guards loose on entire repositories, isolated components, branch diffs, or specific commits. The plugin generates exploit scenarios, assigns severity scores, and even drafts custom patches in an isolated clone of the codebase, ensuring no automated hallucinations overwrite production files without manual human approval.
Installation requires running a single terminal command from Anthropic official marketplace alongside local prerequisites like Claude Code version 2.1.154, Python 3.9.6 or higher, and dynamic workflow support enabled. While the local plugin consumes standard account usage limits, enterprise clients get a separate cloud version currently in public beta for continuous background repository monitoring.
Entrusting code security to an army of artificial intelligence agents might finally end the nightmare of unnoticed backdoors, or simply create the most articulate, confident security bugs humanity has ever witnessed. The days of blaming the intern for unpatched vulnerabilities are officially numbered.
Source: Anthropic
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.