MCP server giving AI agents local file encryption — AES-256-GCM encrypt/decrypt & passphrase generation (free), plus whole-folder encrypt/decrypt (Pro). Keys & files never leave the machine.
VaultCrypt has clear, action-verb naming (encrypt_file, decrypt_file, gen_passphrase) and reasonable descriptions (avg ~100 chars). All 5 tools have input schemas with Zod validation and parameter descriptions. However, output schemas are undocumented, responses are plain text strings with no structured field definitions. Error handling is basic (try/catch with generic messages). No tool annotations (readOnlyHint/destructiveHint). Parameter descriptions lack format constraints (e.g., 'Absolute path' but no validation guidance). Missing idempotency guidance for destructive operations.
Pro: decrypt every .vlt file under a directory (recursively) with one passphrase.
Decrypt a VaultCrypt (.vlt) file with its passphrase.
Pro: encrypt every file under a directory (recursively) to .vlt with one passphrase.
Encrypt a local file with a passphrase (AES-256-GCM, scrypt KDF).
Generate a strong random passphrase for encryption.
Output schemas undocumented. All tools return plain text strings; no structured field definitions for LLM parsing. Agents cannot extract typed data for downstream chaining.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot infer which tools are safe to retry or have side effects.
Error messages lack recovery guidance. 'Error: decryption failed (wrong passphrase or corrupted file)' does not tell the agent what to do next or how to disambiguate.
Parameter descriptions lack format constraints. 'Absolute path to the file' does not specify validation rules, allowed characters, or error cases.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
No confirmation or dry-run pattern for destructive operations. encrypt_file and encrypt_dir immediately overwrite files without agent confirmation.