llm-keys-ui
A plugin for the LLM command-line tool developed by Simon Willison. It provides a web-based interface for API key management, enabling keys to be entered securely through an out-of-band channel to prevent their exposure within the conversational context of AI agents.
Timeline (last 90 days)
Simon Willison released a plugin named llm-keys-ui 0.1, designed to securely configure API keys onto remote machines without pasting them directly into an AI Agent session.
The llm-keys-ui plugin is built on top of Simon Willison's LLM command-line tool ecosystem, reusing the existing key storage mechanism of llm keys and adding a web frontend.
Running the command uvx --with llm-keys-ui llm keys-ui --all launches a web interface for key configuration and returns a URL containing a LAN address or Tailscale device IP.
The llm-keys-ui web interface only displays the names of stored keys (such as anthropic, openai, openrouter, and qwen-dummy); the values of stored keys are never shown.
Using a Tailscale IP as the web interface access address means that only authorized devices belonging to that Tailscale network can open the key configuration page, which is one of the foundational prerequisites for the security of this approach.
The design of llm-keys-ui is essentially to establish an out-of-band configuration channel for credential entry that the agent cannot observe — the human enters keys directly through a browser, while the agent can only read the results via the command line at runtime.
All Facts (6)
The llm-keys-ui plugin is built on top of Simon Willison's LLM command-line tool ecosystem, reusing the existing key storage mechanism of llm keys and adding a web frontend.
50%UnverifiedRunning the command uvx --with llm-keys-ui llm keys-ui --all launches a web interface for key configuration and returns a URL containing a LAN address or Tailscale device IP.
50%UnverifiedThe llm-keys-ui web interface only displays the names of stored keys (such as anthropic, openai, openrouter, and qwen-dummy); the values of stored keys are never shown.
50%UnverifiedUsing a Tailscale IP as the web interface access address means that only authorized devices belonging to that Tailscale network can open the key configuration page, which is one of the foundational prerequisites for the security of this approach.
50%UnverifiedThe design of llm-keys-ui is essentially to establish an out-of-band configuration channel for credential entry that the agent cannot observe — the human enters keys directly through a browser, while the agent can only read the results via the command line at runtime.
50%UnverifiedSimon Willison released a plugin named llm-keys-ui 0.1, designed to securely configure API keys onto remote machines without pasting them directly into an AI Agent session.
50%