Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

With Playwright MCP

Record what an agent did with Playwright MCP, with a trace, screenshots and page snapshots, and share it as a Currents session

Playwright MCP is an MCP server that lets an agent drive a browser. When the agent also records a trace, screenshots and page snapshots, you can check what it did in the browser instead of trusting its summary. Save the recording to Currents as a session, and you get one link to all of it.

What you need

  • Playwright MCP with its tracing tools turned on, so the agent can record a trace. The --caps=devtools option turns them on. For Claude Code:

    claude mcp add playwright -- npx @playwright/mcp@latest --caps=devtools
  • The Currents CLI and a key. See Install the Currents CLI and set a key.

Ask the agent

Reproduce the checkout total bug at http://localhost:3000 with Playwright MCP.
Record it as a Currents session, fix the code, record it again, and post both
links on the pull request.

With a Currents skill, the agent records and shares the session without more instructions. Claude Code recording the bug, sped up:

A terminal: Claude Code empties the traces folder, starts a failed session, starts tracing, fills the discount code and clicks Apply with Playwright MCP, saves a screenshot and a page snapshot, stops tracing, attaches the files captioned before, and shares the session

What you get

A session with the before and after, in the Gallery view:

A session gallery with the before and after screenshots, the before and after traces, the console log and the two page snapshots

The trace opens in a trace player on the session page. A page snapshot is the page's accessibility tree, with its text. Two page snapshots show what changed, even when two screenshots look the same.

Last updated

Was this helpful?