AnyAgent detects your user’s coding agent (Claude Code, Codex, and more) and drives it programmatically. Prompts run under their login and subscription.
import { create, detect } from "anyagent"; const [agent] = await detect(); if (!agent) throw new Error("no coding agent found"); const result = await create(agent).run( "wire our SDK into this project" );
detect() returns whatever coding agent CLI is installed. run() behaves the same across all of them.
You collect no API keys and pay no inference bill.
You ship the prompt and the guardrails; the user’s agent supplies the model and the auth. AnyAgent is for tools that run a defined task against the user’s own project.
Wire your SDK into the user’s codebase from your init command.
Apply a framework upgrade across the user’s repo.
Produce changelogs, tests, or docs from the user’s source.