OpenKeylet

•
136
•

Manage passwords, API keys, SSH credentials, and TOTP codes locally for AI agents via MCP while keeping plaintext secrets out of model context.

OpenKeylet is a local-first credential vault and MCP server designed to let AI coding agents and assistants execute authenticated workflows without exposing sensitive plaintext in conversation contexts. Built in Rust and running locally without cloud dependencies, it securely stores logins, SSH keys, API tokens, and passkeys using XChaCha20-Poly1305 encryption. Agents can request actions—such as browser field autofill, SFTP file transfers, SSH command execution, or authenticated HTTP API requests—while OpenKeylet performs the credential injection on the local device, ensuring secrets remain hidden and workflows stay secure.

Key Features

01Bundled Chromium browser extension for secure credential autofill and passkey management
02Agent-driven actions including SSH command execution, SFTP file transfers, and authenticated HTTP requests
03Encrypted local vault supporting logins, SSH keys, API tokens, passkeys, and TOTP codes
04134 GitHub stars
05Context-safe secret injection via local stdio MCP server without returning plaintext to LLMs
06Automatic detection and configuration repair for clients like Claude Code, Cursor, and Codex

Use Cases

01Enabling automated browser test or scraping workflows requiring secure login authentication and passkeys
02Allowing AI coding agents to deploy code or verify services via SSH without exposing private keys in prompts
03Running agentic workflows that query third-party APIs using Bearer tokens, Basic auth, or secret placeholders safely