"author:vdoflip" matching MCP servers:

Enables AI agents to use individual allowlisted 1Password items by running pre-approved commands one at a time, with each call requiring explicit user approval. Secrets are injected only into the child process environment and masked from the agent, which receives just exit code and output.