Agent-readable docs index: /docs/llms.txt. Full docs in one file: /docs/llms-full.txt. Download /docs/docs.zip to grep all markdown files locally.

Introduction

A web action layer for AI agents. An agent asks for what it wants to do (find), reads a page (read), and runs a typed action (act) that is checked against postconditions. Risky writes wait for a person (approve). Everything is recorded in a hash-chained ledger (explain, verify).
The same seven operations exist in REST, the TypeScript SDK, the Python SDK, the CLI and the MCP server.

the pieces

  • Registry: sites, their Action Manifests and signed recipes.
  • Runtime: runs a recipe, fastest tier first: api, webmcp, replay, dom, agent.
  • Trust plane: policy, approvals, quarantined extraction, credentials kept out of model context, the ledger.
  • Search: pages from providers you configure, actions from the registry, ranked together.

status

Early. Nothing is published to npm, PyPI or a package registry yet; run the packages from the repository. The API shape is in the OpenAPI document.