Tamash — WebMCP Tester

Tracking since Apr 16, 2026.

Changelog

• Jun 5, 2026
  description

  Tamash — AI Web Agent Bridge
  
  Tamash connects your active Chrome tab to AI assistants so they can read and interact with any website without writing a single line of test code or selector.
  
  How it works
  
  When you open a page, Tamash automatically scans it and turns its interactive controls into callable tools. It supports the WebMCP standard, so if a website declares its own tools via the WebMCP API, Tamash discovers and exposes those first — giving AI assistants a semantic, app-aware interface rather than a raw DOM scan.
  
  MCP Mode — use with your AI IDE
  
  In MCP Mode, Tamash runs a local Model Context Protocol server. Any MCP-compatible AI IDE — VS Code Copilot, Cursor, or Claude Desktop — can connect to it and call page tools directly. The connection happens over a local WebSocket; nothing leaves your machine.
  
  Agent Mode — built-in AI agent
  
  In Agent Mode, Tamash runs an AI agent from within the extension sidebar. Connect your own OpenAI, Anthropic, Google, or Ollama account, describe a goal in plain language, and the agent navigates pages, fills forms, and completes multi-step tasks while logging every tool call so you can follow along.
  
  Tool Tester
  
  The Tool Tester tab shows every tool discovered on the current page. You can inspect each tool's input schema and run it manually — useful for verifying what the AI sees before handing off to an agent or IDE.
  
  Remote MCP Servers
  
  In Agent Mode you can connect Tamash to remote MCP servers such as Jira and GitHub, allowing the agent to combine live page actions with external system operations in a single run.
  
  All credentials and tokens are stored locally on your device and are never synced or transmitted to Tamash servers.

  TAMASH — WebMCP Tester
  
  Postman for WebMCP: inspect, test, and orchestrate WebMCP tools exposed by web applications — all from a Chrome side panel.
  
  WebMCP is Chrome's emerging API that lets web apps declare named, callable tools for AI agents (via document.modelContext.registerTool). TAMASH is the first dedicated testing and automation tool built around it.
  
  ──────────────────────────────
  WHAT IS WEBMCP?
  ──────────────────────────────
  
  WebMCP lets a web application declare its own tools — like add_item_to_cart, user_login, or place_order — that AI agents and automated flows can call directly. Instead of fragile CSS selectors or DOM scraping, the app exposes a stable, semantic API that survives UI redesigns.
  
  TAMASH requires Chrome 146 or later with Experimental Web Platform features enabled (chrome://flags/#enable-experimental-web-platform-features).
  
  
  ──────────────────────────────
  FOUR TABS — ONE SIDE PANEL
  ──────────────────────────────
  
  🔧 TOOL TESTER
  Browse every WebMCP tool the current page has registered. Click a tool, fill in parameters, click Run, and see the exact result. No AI needed. Ideal for developers building and validating WebMCP tools before connecting to an agent.
  
  🤖 AGENT
  Describe a goal in plain English and let the built-in AI agent drive the page using its WebMCP tools — turn by turn, up to 25 turns. Supports OpenAI, Anthropic (Claude), Google Gemini, and Ollama (cloud or local). No external server required — the agent runs entirely inside the extension.
  
  🔀 FLOW
  Design reusable sequences of WebMCP tool calls that replay on demand — no AI prompt needed at run time. Create flows manually, or click "Save as Flow" after any successful agent run to capture every tool call automatically. Import and export flows as plain JSON to share with your team. Drag steps to reorder, edit parameters inline, and run with a single click.
  
  ⚙ SETTINGS
  Configure your AI provider, select a model, and save your API key. Also where you connect Remote MCP servers — add Jira or GitHub so their tools appear in the agent alongside your page's WebMCP tools.
  
  
  ──────────────────────────────
  TRY IT WITH VEGCART
  ──────────────────────────────
  
  VegCart is a purpose-built WebMCP demo app — a fresh produce e-commerce store with 16 registered WebMCP tools covering products, cart, checkout, wishlist, user accounts, and order history.
  
  Open https://vibetestq.com/testweb/vegcart/ in Chrome, click the TAMASH icon, and the Tool Tester immediately shows all 16 tools. Switch to the Agent tab and type:
  
    "Login as [email protected] / test123, add Alphonso Mango ×2 and Spinach ×3, place an order to Bengaluru 560001, then show me the order history"
  
  Watch the agent execute every step — login, clear cart, add items, checkout, verify — in one run.
  
  12 pre-built VegCart flows are available to download and import at:
  https://vibetestq.com/tamash/help#vegcart-flows
  
  
  ──────────────────────────────
  AI PROVIDERS SUPPORTED
  ──────────────────────────────
  
  - OpenAI: gpt-5.4, gpt-5.4-mini, gpt-5.4-nano
  - Anthropic: claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5
  - Google: gemini-2.5-pro, gemini-2.5-flash, gemini-3.1-pro, gemini-3.1-flash
  - Ollama: qwen3.5, qwen3-coder-next, deepseek-v3.2, and more
    (Use https://ollama.com cloud or override Base URL to http://localhost:11434 for local Ollama — no API key required for local)
  
  Your API key is stored locally on this device only (chrome.storage.local) and is never synced or sent to VibeTestQ.
  
  ──────────────────────────────
  HOW IT WORKS
  ──────────────────────────────
  
  1. Install the extension.
  2. Navigate to chrome://flags, enable "Experimental Web Platform features", and relaunch Chrome.
  3. Open a WebMCP-enabled app (try VegCart at the link above).
  4. Click the TAMASH icon to open the side panel.
  5. Use Tool Tester to inspect and call tools manually, Agent to run AI-powered workflows, or Flow to replay saved sequences.
  
  
  ──────────────────────────────
  PRIVACY
  ──────────────────────────────
  
  - No data is sent to VibeTestQ
  - No analytics or telemetry
  - API keys and credentials are stored in chrome.storage.local only — never synced
  - Data is transmitted to third-party AI providers or Remote MCP endpoints only when you explicitly use those features
  - The extension does not read cookies, localStorage, browsing history, or DOM content beyond WebMCP tool responses
  
  Full privacy policy: https://vibetestq.com/tamash/privacy
  
  
  ──────────────────────────────
  REQUIREMENTS
  ──────────────────────────────
  
  - Chrome 146 or later
  - "Experimental Web Platform features" enabled at chrome://flags
  - A WebMCP-enabled web application (VegCart is provided as a ready-to-use demo)
  - An AI API key (OpenAI, Anthropic, Google, or Ollama) for Agent tab features
    (Tool Tester and Flow tabs work without any API key)
  
  
  ──────────────────────────────
  LINKS
  ──────────────────────────────
  
  Home: https://vibetestq.com/tamash/
  User Guide & Flows: https://vibetestq.com/tamash/help
  VegCart Demo: https://vibetestq.com/testweb/vegcart/
  Privacy Policy: https://vibetestq.com/tamash/privacy

• Jun 5, 2026
  short_description

  Makes web apps AI-agent-ready with Agent Mode, WebMCP, DOM automation, and Jira/GitHub Remote MCP for MCP clients.

  Postman for WebMCP: inspect, test, and run WebMCP tools. Built-in AI agent, flow designer, and Jira/GitHub Remote MCP.

• Jun 5, 2026
  name

  Tamash - AI Web Agent Bridge

  Tamash — WebMCP Tester

• Jun 5, 2026
  host_permissions

  <all_urls>, https://api.openai.com/*, https://api.anthropic.com/*, https://generativelanguage.googleapis.com/*, https://ollama.com/*

  <all_urls>, https://api.openai.com/*, https://api.anthropic.com/*, https://generativelanguage.googleapis.com/*, https://ollama.com/*, https://mcp.atlassian.com/*, https://api.githubcopilot.com/*

• Jun 5, 2026
  permissions

  activeTab, scripting, offscreen, alarms, storage, sidePanel, tabs, cookies

  activeTab, scripting, storage, sidePanel, tabs

• Apr 29, 2026
  description

  Tamash — AI Web Agent Bridge
  
  Connect your browser to AI assistants like GitHub Copilot, Cursor, and Claude Desktop.
  
  Let them read, navigate, and act on any website—directly from your IDE chat.
  
  🚀 What it does
  
  Tamash turns web pages into structured tools your AI can use.
  
  Your AI can:
  
  Search and filter content
  Click buttons and submit forms
  Navigate flows
  Extract structured data
  
  No automation scripts. No selectors. No setup.
  
  ⚙️ How it works
  Detects WebMCP tools exposed by websites
  Connects them to your AI via MCP
  Executes actions directly in your browser
  🧠 Use cases
  Automate repetitive browser tasks
  Test web apps using natural language
  Build AI-driven workflows
  Control internal tools with AI
  🧩 Works with
  VS Code + GitHub Copilot
  Cursor
  Claude Desktop
  Any MCP-compatible AI client
  ⚡ Quick setup
  Install the extension
  
  Add to MCP config:
  
  npx tamash-wmcp
  Open a WebMCP-enabled site
  Start chatting with your AI
  🔒 Privacy
  Runs entirely in your browser
  No data collection
  No external servers

  Tamash — AI Web Agent Bridge
  
  Tamash connects your active Chrome tab to AI assistants so they can read and interact with any website without writing a single line of test code or selector.
  
  How it works
  
  When you open a page, Tamash automatically scans it and turns its interactive controls into callable tools. It supports the WebMCP standard, so if a website declares its own tools via the WebMCP API, Tamash discovers and exposes those first — giving AI assistants a semantic, app-aware interface rather than a raw DOM scan.
  
  MCP Mode — use with your AI IDE
  
  In MCP Mode, Tamash runs a local Model Context Protocol server. Any MCP-compatible AI IDE — VS Code Copilot, Cursor, or Claude Desktop — can connect to it and call page tools directly. The connection happens over a local WebSocket; nothing leaves your machine.
  
  Agent Mode — built-in AI agent
  
  In Agent Mode, Tamash runs an AI agent from within the extension sidebar. Connect your own OpenAI, Anthropic, Google, or Ollama account, describe a goal in plain language, and the agent navigates pages, fills forms, and completes multi-step tasks while logging every tool call so you can follow along.
  
  Tool Tester
  
  The Tool Tester tab shows every tool discovered on the current page. You can inspect each tool's input schema and run it manually — useful for verifying what the AI sees before handing off to an agent or IDE.
  
  Remote MCP Servers
  
  In Agent Mode you can connect Tamash to remote MCP servers such as Jira and GitHub, allowing the agent to combine live page actions with external system operations in a single run.
  
  All credentials and tokens are stored locally on your device and are never synced or transmitted to Tamash servers.

• Apr 29, 2026
  short_description

  Makes any web app AI-agent-ready. Bridges tabs to VS Code Copilot, Cursor & Claude Desktop via MCP. Supports WebMCP and DOM tools.

  Makes web apps AI-agent-ready with Agent Mode, WebMCP, DOM automation, and Jira/GitHub Remote MCP for MCP clients.

• Apr 29, 2026
  host_permissions

  <all_urls>

  <all_urls>, https://api.openai.com/*, https://api.anthropic.com/*, https://generativelanguage.googleapis.com/*, https://ollama.com/*

• Apr 29, 2026
  permissions

  activeTab, scripting, offscreen, alarms

  activeTab, scripting, offscreen, alarms, storage, sidePanel, tabs, cookies

Permissions & access

Permissions

activeTabscriptingstoragesidePaneltabs

Host access

<all_urls>, https://api.openai.com/*, https://api.anthropic.com/*, https://generativelanguage.googleapis.com/*, https://ollama.com/*, https://mcp.atlassian.com/*, https://api.githubcopilot.com/*

Screenshots