Free IDE extension for risk-free vibe coding — keep secrets out of AI.
Cursor Security
Cursor reads your open files, recent edits, and project context to power its AI commands. That same context can contain API keys, database credentials, patient data, or other sensitive content that should never leave your machine. SoterAI IDE Guard runs locally in VS Code-compatible editors including Cursor, scanning your workspace for secrets, PII, prompt-injection text, and risky MCP tool configurations before they reach the AI model.
Runs locally in your editor. Secret, PII, prompt-injection, and MCP scanning happen on your machine before anything reaches an AI model.
Scan files, selections, or your entire workspace for API keys, tokens, database URLs, Aadhaar patterns, PAN, and other sensitive data — entirely on your machine before anything reaches Cursor's AI.
Use 'Scan Before AI Prompt' to review and redact context immediately before sending it to Claude, GPT-4, or any model inside Cursor. See exactly what the AI will read.
Cursor supports MCP tools. SoterAI IDE Guard scans your MCP server configurations and flags over-broad tool permissions before you enable them in your Cursor workspace.
Detect injected instructions in files, repository content, or pasted snippets before they enter a Cursor AI command — including invisible unicode character tricks.
Keep a local ledger of what context Cursor AI saw in your session so you can audit exposure after the fact — without sending any data to a server.
Warn on destructive or exfiltration-prone shell commands suggested by Cursor's terminal AI before they run on your machine.
Install SoterAI IDE Guard from the VS Code Marketplace
SoterAI IDE Guard is built on the VS Code extension API and works in Cursor, which is built on the same platform. Install from the VS Code Marketplace or the Open VSX Registry.
Open your Cursor workspace
The extension activates automatically. All scanning defaults to local — no data is sent anywhere without your explicit opt-in.
Scan before you run a Cursor AI command
Use the 'Scan Before AI Prompt' command from the Command Palette to review your context for secrets, PII, and injection text before sending it to Cursor's AI.
Review and redact
Redact detected sensitive spans directly in the scan results. The redacted version stays on your machine; the original file is unchanged.
No security tool is perfect. Here is what this feature does not claim to do, so you can layer defenses appropriately.
Yes. Cursor is built on VS Code's extension API. SoterAI IDE Guard installs and runs in Cursor using the same extension host. Install it from the VS Code Marketplace.
Yes. Use the 'Scan Before AI Prompt' command from the Command Palette to review your current file or selection for secrets, PII, and injection text before running a Cursor AI command.
No. Secret, PII, prompt-injection, and MCP scanning run locally in the extension host on your machine. Cloud features are opt-in and disabled by default.
The extension is model-agnostic and editor-side. It inspects the context you are about to send, regardless of which AI model Cursor uses under the hood.
Install SoterAI IDE Guard and scan secrets, prompts, MCP tools, and terminal commands locally before they ever reach an AI model.
Install for VS Code