STAR Plugin (IDE)

Bright DAST agents and skills for AI coding tools — Cursor, Claude Code, Codex, GitHub Copilot, and Antigravity CLI

STAR Plugin brings Bright's dynamic security testing directly into your AI coding tool. Instead of running in CI as a standalone binary (like STAR), the plugin works interactively inside your IDE — you ask an agent to scan, and it handles the rest.

The plugin ships as a native package for each supported tool, wiring its orchestration agents and skills to the Bright MCP server. You get the same Bright scanning engine, the same vulnerability coverage, but driven conversationally from your editor.

STAR Plugin vs. STAR (CI) vs. Bright MCP

STAR (CI)STAR Plugin (IDE)Bright MCP (raw)
Where it runsYour CI/CD runnerYour IDE / AI coding toolAny MCP-capable client
How you use itTriggered by PR/push/scheduleAsk an agent in chatCall MCP tools directly
Automation levelFully autonomous (scan → fix → PR)Agent-guided, interactiveManual — you orchestrate
Fixes codeYes, commits to branchYes (remediation-loop agent)No — tools only
SetupEnv vars in CIPlugin install + env varsMCP server config
Best forContinuous pipeline securityDev-time security from the editorCustom integrations

How it works

flowchart LR
    subgraph IDE["Your AI Coding Tool"]
        User["You (chat)"] --> Agent["STAR Plugin Agent"]
        Agent --> Skills["Skills<br/>(analyze, auth, register,<br/>scan, fix)"]
    end

    subgraph MCP["Bright MCP Server"]
        Tools["MCP Tools<br/>(runScan, runDiscovery,<br/>createRepeater, …)"]
    end

    subgraph Cloud["Bright Cloud"]
        Scanner["Dynamic Security Scanner"]
    end

    Skills --> Tools
    Tools --> Scanner
    Agent -.->|"private/local targets"| Repeater["Bright Repeater<br/>(on your machine)"]
    Repeater ==>|"outbound only"| Scanner
  1. You ask the agent to scan your app (local, staging, or any authorized target).
  2. The agent analyzes your codebase to discover endpoints and tech stack.
  3. It reaches the target — starting it locally or health-checking a supplied URL.
  4. For private/local targets, it spins up a Bright Repeater on your machine.
  5. It configures authentication if the app requires login.
  6. It registers the attack surface and runs dynamic scans via Bright's engine.
  7. (Remediation loop) It fixes findings, restarts the app, and re-scans to prove the fix.

Supported tools

ToolPackage
Cursorcursor/
Claude Codeclaude-code/
Codexcodex/
GitHub Copilotgithub-copilot/
Antigravity CLIantigravity/
Devinclaude-code/ — Devin reads Claude plugin manifests, so no separate package is needed. See Devin (Desktop & CLI).

All packages wire the same agents and skills — only the tool-specific install mechanics differ.

Requirements

  • BRIGHT_TOKEN — a Bright API token (personal, project, or org-level)
  • BRIGHT_HOSTNAME — your Bright cluster hostname (default: app.brightsec.com)
  • The AI coding tool installed and configured
📘

Project-scoped tokens are recommended

With a project-scoped BRIGHT_TOKEN, the agents automatically use the only reachable project. With an org-wide token, you'll need to name the project when you start — the agents ask rather than guess.

Next steps


What’s Next

Did this page help you?