Skip to content
Dennis Lee edited this page May 27, 2026 · 1 revision

title: rodney radar_quadrant: Tools radar_ring: Assess radar_position: inner

rodney

rodney is a CLI tool by Simon Willison for AI agents to interact with the web in a structured, safe-by-default manner. It fetches URLs and returns clean text or Markdown, follows links, and interacts with forms — providing web browsing capability to agents without exposing raw HTML or enabling unsafe browsing behaviours. JavaScript execution is off by default; cookies do not persist between calls unless explicitly configured.

Radar Assessment

AI agents that need to research the web typically either receive raw HTML (hard to reason about) or use a scraping service without safety constraints. rodney is designed with agent use in mind: its output is structured for LLM consumption, its defaults prevent accidental state accumulation across calls, and its interface maps directly to the operations an agent needs — fetch, follow, extract — rather than full browser simulation.

Placed in Assess at inner position because structured, safe web interaction is a real gap in agent tooling. rodney fills it with a well-considered design by a credible author (Simon Willison), and complements Showboat in a small toolkit for agents that need to research and verify findings from the web. The inner position reflects that developers building agents with web research capabilities should be evaluating rodney now alongside heavier browser automation tools.

The safe-by-default design philosophy — no JavaScript, no persistent cookies, structured output — makes rodney more predictable in agent pipelines than general-purpose browser automation, at the cost of not handling JavaScript-rendered content without explicit configuration.

Clone this wiki locally