Yad Agent Mcp

io.github.holistisPublished by GitHub account holistisv0.1.1Updated Oct 24, 2026

MCP wrapper for Yad, a local browser-automation companion, for coding agents.

VerifiedSTDIODesktop onlyBrowser AutomationDeveloper Tools

Overview

AI-generated overview

A local MCP wrapper that lets a coding agent drive Yad, a privacy-first browser companion, inside the user's own logged-in browser.

What you can do
  • Check whether Yad's companion is running and connected to the browser
  • Open a URL in the browser tab Yad controls
  • Read the current page's URL, title, visible text and links
  • Hand Yad a plain-language browser task and wait for the result
  • Re-read the result of the most recent task
When to use it
Use it when you want an agent to work inside your real, already-logged-in Chrome, Edge or Brave session — navigating, reading pages, or performing plain-language browser tasks — on a machine where Yad is already installed.
Requirements
Runs locally over stdio via npm (yad-agent-mcp) and needs Node.js 20+. Yad must be installed and running: a Chrome, Edge or Brave extension plus the local companion app. The companion's HTTP API is reached at 127.0.0.1:3747 by default; set YAD_COMPANION_URL if it uses another port. No authentication is declared.
Before you install
It drives your real logged-in browser session, so an agent can act with your existing sessions and cookies. Tasks can click and navigate on live sites. Yad itself blocks checkout, payment and order placement without your own confirmation, and this wrapper neither adds nor removes that limit. It only works locally and is not a remote server.

Installation

In SourceWeft

  1. Open Yad Agent Mcp in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

yad-agent-mcp

An MCP server that lets an AI agent drive Yad, a local, privacy-first browser-automation companion, in the user's real, already-logged-in browser.

This is a thin wrapper: it does not run a browser itself. It talks to Yad's existing local HTTP API on 127.0.0.1:3747 over stdio, so any MCP-compatible client (Claude Code, Cursor, Claude Desktop, etc.) can ask Yad to navigate, read a page, or carry out a plain-language browser task, and get the result back.

Local-only by design. This is not a hosted or remote MCP server. It only works on a machine that already has Yad installed and running (extension in Chrome/Edge/Brave, companion app active). If you're looking for a server your agent can call over the internet with zero local setup, this isn't that, and honestly, nothing that drives a user's real logged-in browser session should be.

Requirements

  • Yad installed and running (Chrome, Edge, or Brave extension plus the local companion app)
  • Node.js 20+

Install

bash
npm install -g yad-agent-mcp

Configure your MCP client

Claude Code (.mcp.json or claude mcp add):

json
{  "mcpServers": {    "yad-agent": {      "command": "yad-agent-mcp"    }  }}

If Yad's companion runs on a non-default port, set YAD_COMPANION_URL (defaults to http://127.0.0.1:3747).

Tools

ToolWhat it does
yad_statusChecks whether Yad's companion is running and connected to the browser. Call this first.
yad_navigateOpens a URL in the browser tab Yad controls.
yad_captureReads the URL, title, visible text, and links of the current page.
yad_run_goalGives Yad a plain-language task ("find the price of the first item", "click the login button") and waits for the result.
yad_last_resultRe-reads the result of the most recent yad_run_goal call.

Yad cannot check out, pay, or place an order without the user's own confirmation, regardless of what an agent asks it to do. This wrapper doesn't add or remove that limit, it's enforced by Yad itself.

Source

Part of the al-yad monorepo. Issues and PRs welcome there.

License

MIT

Source: packages/mcp-server/README.md at commit 552a81a

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.1.1LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts