Team
1–1,000 employees
$5 per employee / month, billed yearly
$6 billed monthly
Browser and IDE protection, the security dashboard, and weekly reports.
PromptFlowEngine Enterprise
Keep secrets out of AI prompts, without reading anyone’s prompts
Warn, redact, or block credentials and personal data before they are sent to AI chat and coding agents. Your security team sees what kind of data, where, and what happened. Never the prompt.
How it works
- Verify your domain with a DNS record, then employees connect with their work email, or IT deploys an enrollment key through Chrome policy or VS Code settings.
- Set a policy per kind of data: allow, warn, redact, or block. Scanning runs on the employee’s device with the same deterministic engine as the free tools.
- See the results on the dashboard, in weekly and monthly reports, and in your own SIEM or API through signed webhooks.
What is covered, and what is not
Pre-submit blocking only exists where a tool lets us run before it sends. We verified each tool’s documented mechanism; where none exists, we say so.
| Tool | Mechanism | Warn | Block | Unsent drafts | Notes |
|---|---|---|---|---|---|
| ChatGPT, Claude, Gemini, Grok, Perplexity, DeepSeek | Chrome extension | Yes | Yes | Yes | Enter key, send button, and form submit. Voice input and file uploads are not checked. |
| Claude Code | UserPromptSubmit hook | Yes | Yes | No | Blocked prompts are erased from the session. |
| Cursor | beforeSubmitPrompt hook | No | Yes | No | No warning channel: warn-level findings are recorded only. Cursor reports occasional hook failures; treat as best effort. |
| Windsurf (incl. its JetBrains plugin) | pre_user_prompt hook | No | Yes | No | No warning channel. |
| GitHub Copilot Chat in VS Code | VS Code agent hooks (Preview) | Yes | Partly | No | Agent sessions only; ask mode is not documented to run hooks. |
| Gemini CLI / Gemini Code Assist agent mode | BeforeAgent hook | No | Partly | No | Gemini CLI documented; Code Assist only where it runs Gemini CLI. |
| Continue, other VS Code assistants | Open-file exposure warnings | Partly | No | Yes | No pre-submit hook exists; we warn when an open file holds credentials an assistant could attach. |
| JetBrains AI Assistant | None | No | No | No | JetBrains offers no third-party interception API. |
Privacy-first by construction
- Prompts are scanned on the device. The event format has no text field: the server rejects any event that carries one.
- People appear as a keyed pseudonym per organization, not by name or email.
- Events are kept 30 days and daily totals 400 days by default; you choose both.
- Delivery to your API is HTTPS-only and HMAC-signed, with buffered retries for about 37 hours.
- Role-based access, an audit log of every admin action and report download, and one-click export or deletion of all organization data.
Details: how Enterprise handles data and the Enterprise terms.
Pricing
Per employee covered. Your band is set by the number of employees on the plan.
Business
1,001–10,000 employees
$3.75 per employee / month, billed yearly
$4.50 billed monthly
Everything in Team, plus real-time delivery to your API and monthly reports.
Enterprise
10,001–100,000 employees
$2.50 per employee / month, billed yearly
$3 billed monthly
Everything in Business, plus custom retention and a named security contact.
More than 100,000 employees, or need an invoice: contact sales. Prices in USD, excluding taxes.
Compliance status
PromptFlowEngine is not yet certified under ISO/IEC 27001 or any other scheme. We have implemented the controls described above and are preparing an ISO 27001 information security management system, and we will say so here only once certification is issued. Our data processing agreement for GDPR purposes is in legal review; ask us for the current draft.