📰 Story

Show HN: Agent-data – a CLI for giving agents real-time, structured data

Hey HN, Justin here, and we’re building agent-data: a CLI tool that gives AI agents access to real-time, structured data without requiring them to use browser automation. The idea came from trying to get agents (e.g., OpenClaw, Hermes Agent) to perform more useful tasks than simply organizing our calendars or monitoring our emails. We wanted them to answer questions and take actions that depend on timely external data, e.g., “Can you find me roundtrip flight options from NYC to SF next month”. Personal assistant agents appear to be great frameworks for maintaining user preferences and acting with that information (e.g., “Justin hates early morning flights,” “Justin doesn’t mind a layover if it’s shorter than 3 hours,” etc), but existing tools for providing assistants with actionable, external data are surprisingly limited. For example, browser automation can work, but it’s generally slow, expensive, and flaky [1]. Ad-hoc scraping works well in narrow cases, but can become quite expensive at scale. MCP servers can inject tool schemas into context on every turn (inflating token consumption), and are often either too granular or too coarse in their design. So we built agent-data around a simpler interface: give the agent the ability to discover the right endpoint at runtime, call it via CLI, and get back compact structured output. For example, to answer my earlier question, an agent might use: `agent-data call search-flights --origin_iata JFK --destination_iata SFO [--params]` W