Skip to content

Security: retyc/retyc-cli

Security

SECURITY.md

Security Policy

Reporting a Vulnerability

Please do not open a public GitHub issue for security vulnerabilities.

See our security policy and contact details at: https://retyc.com/.well-known/security.txt


Known Limitations (MCP server mode)

Passphrase exposed as an environment variable RETYC_KEY_PASSPHRASE is passed to the MCP server via an environment variable. On some systems, environment variables of a running process may be readable by other processes belonging to the same user. Some MCP clients may also log or surface environment variables in their UI or debug output.

The LLM can read metadata, never file content Tool responses may include transfer and dataroom metadata (filenames, sizes, timestamps, member lists, etc.) in plaintext - this data passes through the MCP client and the language model. File content, however, is end-to-end encrypted: it is decrypted locally and written directly to disk, and never included in tool responses.

Transfer passphrases pass through the LLM Some transfers are protected by a per-transfer passphrase. If a tool requires one, the LLM may ask the user to provide it; that passphrase then transits the MCP client and the model context. Do not use a transfer passphrase that is reused elsewhere. To avoid this entirely, prefer key-based transfers by specifying the recipient's email address: the transfer is then encrypted to their public key and no passphrase is involved.

Security depends on MCP client isolation The MCP server runs with full access to the Retyc token and passphrase. The confidentiality of these credentials depends on the isolation guarantees of the MCP client in use.

There aren't any published security advisories