mcp-warden

mcp-warden

Local-first credential broker that keeps secrets out of AI agent context

Visit mcp-warden

About mcp-warden

mcp-warden is a local-first credential broker for the Model Context Protocol (MCP). It allows AI coding agents (Claude Code, Cursor, Codex, etc.) to call HTTP APIs, SSH, and databases through scoped presets, with secrets stored in the OS keychain and never exposed to the agent's context. It enforces method/path allow-listing, approval controls, and prompt-injection defenses, and supports a Claude Code plugin for seamless integration.

Resources

Product Website

Visit mcp-warden's official website for product details and getting started.

Visit website →

Documentation

Comprehensive guides and API references for using mcp-warden effectively.

View docs →

Issues

Community discussions and troubleshooting related to mcp-warden.

Join the community →

Releases

Changelog and release notes for mcp-warden updates and versions.

See releases →