Trust

Security

Last updated 2026-11-08

Sendmind connects to the tools you code with, so we treat security as part of the product, not an add-on.

A firewall for your agent

Anything you send us is treated as information, never as instructions. Captions, transcripts and emails can contain hidden prompt injection ("ignore previous instructions…"). Sendmind detects these, stores them as quoted text and flags them, so a post you saved can never take control of your agent.

You approve every rule

Learnings live in their own folder that your agent can read. Nothing is written into your CLAUDE.md, Cursor rules or any instruction file without your explicit approval.

Local recall

Your learnings sync to a small database on your computer, and searching them happens there. Your prompts are not sent to our servers for recall.

Data handling

• Encryption in transit (TLS) everywhere.
• Media files are deleted as soon as processing finishes.
• Access to production systems is limited and logged.
• We never sell data and never train AI models on your content.

Website protections

This site uses strict security headers (Content Security Policy, HSTS, frame blocking), bot protection on forms, and input validation on every endpoint.

Report a vulnerability

Found something? Email contact@sendmind.dev with details. Please give us reasonable time to fix it before sharing it publicly. We read every report. Our security.txt has the same contact.