A naming convention that lets AI agents understand your data without being told what it means.
Have your coding agent read the Agent Skill and apply the convention across your code.
An agent’s hands and eyes on the world.
An agent can think, but to do anything it needs tools — to fetch a page, run a query, move a payment. The tools that already exist were built for people: their output is text meant for a human to read, and a failure is a sentence a human is supposed to interpret. An agent ends up scraping that text and guessing. Agent-First Kit builds the tools an agent actually needs — some are the essentials rebuilt the right way, some are entirely new. You run them like any command-line tool, but everything they return is structured: every value comes with its meaning attached, and every failure carries a stable, named code instead of a sentence. A small kit today, and growing.
Give an AI agent any URL and get back a usable page — fetched directly, or rendered in a real browser when the page needs one — with a human able to take over the same browser for a login, captcha, or 2FA.
Hand it to your agent and it fetches any URL as data — a plain request, or the browser engine the site demands.
A PostgreSQL interface for AI agents: reliable, structured, explicit, and read-only by default.
Hand it to your agent and it queries PostgreSQL through structured output, in place of psql.
A payment tool for AI agents — send and receive across five networks through one interface, with spending limits you control.
Hand it to your agent and it moves money across five networks, under the spending limits you set.