Secretctl
A free secrets manager built for the AI-coding-assistant era, designed so tools like Claude Code can use your API keys without ever actually reading the plaintext value.
🔗 Visit Secretctl
Description
When an AI coding assistant runs a command that needs an API key, the usual approach is to just have that key sitting in a .env file the assistant can read directly — which means every prompt, log, and chat transcript is a potential place for that secret to leak. Secretctl exists to close that gap: it injects the real secret only into the running process, while the AI agent itself only ever sees a sanitized, redacted version of any output.
Secretctl is a local-first, single-binary secrets manager available as both a CLI and a desktop app, with AES-256-GCM encryption and Argon2id key derivation. It integrates directly with Claude Code via MCP, so an agent can trigger a command that needs a secret without that secret ever entering the agent's context window or memory. It includes tamper-evident audit logging (HMAC-chained), password generation, backup/restore, and import/export, and stores everything locally with no cloud dependency. It's free and open-source under Apache 2.0.
💬 Our review
The short version: Secretctl is a small but well-thought-out answer to a very 2026 problem — how do you let an AI coding agent actually use your API keys without handing it the keys themselves.
The MCP integration is the practical core of the design: rather than relying on the agent to 'be careful' with a secret it can technically read, Secretctl structurally prevents the plaintext from ever reaching the agent's context, which is a fundamentally stronger guarantee than convention or prompting. The tamper-evident audit log is also a nice touch for anyone who wants to verify after the fact that nothing leaked.
The honest caveat: it's an early, small project (single-digit GitHub stars at last check), built by one maintainer, so the usual early-project risks apply — less battle-testing, slower response to edge cases. For developers specifically worried about AI agents touching their secrets, it's a focused, purpose-built tool; for general password management unrelated to AI workflows, a mature option like Bitwarden remains the safer default.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Entièrement gratuit et open-source (Apache 2.0), local-first, aucun cloud requis.
Pros
Intégration MCP native — les secrets n'entrent jamais dans le contexte de l'agent
Chiffrement AES-256-GCM + Argon2id, tout en local
Journal d'audit inviolable (chaîné HMAC)
Application CLI et desktop, binaire unique, cross-plateforme
Cons
Projet très jeune et petit (mono-mainteneur, peu d'étoiles GitHub)
Pas conçu comme gestionnaire de mots de passe généraliste
Encore peu éprouvé à grande échelle