Skip to main content

Yam documentation

Yam is a deterministic automation runtime with a standard agent surface, from Svatah Labs. Describe a behaviour once in plain language, bind it to the real elements by driving the real platform, then replay that plan deterministically, with no model in the loop. The support matrix says which platforms that covers and how far each has been driven; it is generated from measured runs rather than written.

The documentation is organised by what you are trying to do.

If you want toRead
connect an AI agent to YamYam MCP
install it and run somethingSetup
see everything Yam doesFeatures
learn from working examplesExamples
look up a command, tool or endpointAPI reference
get bindings into an existing Playwright suitePlaywright quick start
do one specific taskGuides
understand how and why it worksConcepts
work on Yam itselfDeveloper guide
send a changeContributing

Start here​

  1. Setup, from npm install to a working machine, including the permissions macOS wants and which program it grants them to.
  2. Features, what Yam does, one feature at a time, with pictures.
  3. Examples, working code for the things people do most.

Getting started, in detail​

  1. Install Yam and control something, from npm install to a browser and a native application driven by hand, then by an agent through Yam MCP.
  2. Bindings in a plain Playwright project, the smaller thing to adopt if you already have a Playwright suite.
  3. Your first flow, from yam init to a replayed plan.
  4. One plan, three ways to run it, as a test, a workflow and an agent tool.

Guides​

Concepts​

Reference​

Written by hand:

Generated from the code by pnpm docs:build, and checked in CI so they cannot drift:

Project​