Skip to content

README.md

76 lines · 2.6 KB · Markdown

Playrunner MCP

Run Playwright test suites on Playrunner Cloud from Kody — build a workflow, start a run, and read what came back, without leaving your agent.

Intent

Playwright runs tests. Somebody still has to run everything around them — compute, environments, credentials, schedules, artefacts and reports. Playrunner owns that orchestration, and this package puts it in front of a Kody agent.

The package exists to make connecting trivial. Installing it and calling the root export registers the Playrunner MCP server on your account; you approve access in the browser once, and the tools become available as kody.mcp["playrunner"].*.

Success looks like asking your agent to run a suite against staging and getting back a real result with a report URL.

Prerequisites

A Playrunner Cloud account. The free beta needs no card, and you sign in with GitHub or Google.

You do not need an API token. Playrunner supports dynamic client registration, so Kody registers itself and you approve in the browser.

Setup

  1. Install the package.
  2. Ask your agent to connect Playrunner, or call the root export directly.
  3. Open the authorisation link it returns and approve access.
  4. Ask for status() to confirm the tools were discovered.

How you know it is working

status() reports connected: true and lists the tools. From there:

const workflows = await kody.mcp['playrunner'].list_workflows({})

Try it without writing a workflow first

A new Playrunner account has nothing to run, so ask your agent to build something against the public demo suite:

Create a Playrunner workflow that runs the tests in
playrunner/playwright-demo-testsuite on the demo/sharding-report-merge branch,
with auto sharding. Then run it and tell me what failed.

That branch holds 128 tests, so auto sharding splits them across runners and merges the reports into one result. 1.61.1/pass and 1.61.1/fail are there too, for a clean run and a deliberate failure.

What your agent can do

Discover projects and workflows, author them, connect a GitHub repository, run a saved workflow, and inspect the results. Authenticated tests are supported through Playrunner's Authentication Profiles.

Runs started this way consume the same workflow-run allowance as any other trigger.

Scoping access

Set Usage on /account/mcp-servers/:id to limit which packages may call Playrunner, and revoke access from your Playrunner Cloud account at any time.

Documentation