OpenTiny WebMCP SDK CLI
webmcp-cli from OpenTiny drives a Chrome page through MCP tools, and a skill from the set teaches an agent to use it.
- 125 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent drives a real Chrome window with clicks, input and the sessions of open sites.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/opentiny-webmcp-sdkDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent drives a real Chrome window with clicks, input and the sessions of open sites.
This is third-party code. Review the repository files before installing.
What it does
OpenTiny NEXT-SDKs implements WebMCP: a page describes its actions as tools and an agent calls them. For agents the toolkit has two artifacts. One is `webmcp-cli`, which runs Chrome in the background through Puppeteer and turns clicks, input and scrolling into MCP tool calls, and the other is the `packages/webmcp-skill` directory of skills. The agent opens a page with `webmcp-cli tabs open`, reads state with an indexed DOM through `webmcp-cli state` and acts through `webmcp-cli run`. The skill reads `SKILL.md`, and sub-skills such as the Excalidraw one hint at command formats on a specific site. The `document.modelContext` polyfill also works on pages without source changes.
Who it is for. Front-end developers and testers who want an agent to drive a web application without a separate driver for each site.
Good fit when
- An agent draws a diagram in Excalidraw by calling a page tool.
- A scenario in a web application has to be checked by driving buttons and fields by DOM index.
- Your own application should be opened to AI control through `registerTool` with minimal code changes.
- An internal site with no API has to be walked by an agent through click and input steps.
Not a fit when
- Scraping without a browser window or a ready Playwright MCP is enough.
- Chrome has work accounts open that an agent must not touch.
- You need a supported production service: native WebMCP in browsers is still experimental.
Example request
Open excalidraw.com, show the page state and draw a rectangle labeled "entry".Limitations
Native WebMCP in browsers is still experimental, and by default the SDK forces its own polyfill: the native version can crash the renderer on a `getTools()` call. The agent drives a real Chrome, so use a separate profile without work accounts. Install commands and the full skill list are kept in the project documentation. Most of the README covers the developer SDK, while the CLI has a separate section.
How to disable. Remove the webmcp-skill skill from the agent directory and the package with `npm uninstall -g @opentiny/webmcp-cli`. Close the Chrome window the command opened.
FAQ
What is WebMCP?
An extension of the MCP protocol for browsers: a page hands tools and resources to an agent through `document.modelContext`. While the native API is experimental, the SDK adds a polyfill.
How does the agent see a page?
`webmcp-cli state` returns the address, title, tabs, the MCP tool list and a DOM tree with indexes such as `[18]<button>`. Then `webmcp-cli run page-agent-tool` clicks by index or types text.
Do I have to change the site code?
No, the polyfill picks up third-party pages without edits. For your own applications, tools are declared with `registerTool`, giving exact actions instead of DOM clicks.
What is inside webmcp-skill?
The `packages/webmcp-skill` directory with `SKILL.md` and sub-skills. One covers drawing in Excalidraw, another searching on Baidu.
Which browser is needed?
Chrome: `webmcp-cli` launches it in the background through Puppeteer. A page is opened with `webmcp-cli tabs open` and the target address.
Discussion
No questions or reviews yet.
Related
CLI toolsAn MCP server and Chrome extension that let the agent drive your already open browser with your logged-in sessions, all locally
Local web search for agents: searches, reads pages and returns compact passages with sources, saving context tokens
A CLI, Chrome extension, SDKs and MCP server that let an agent drive your real Chrome, whatever the agent runtime
The Chrome DevTools team's official MCP server: the agent drives a live Chrome, reads network and console, and records performance traces
A Chrome extension with an MCP server: the agent drives your everyday browser with its sessions, tabs, bookmarks and history
Console browser mb for agents: open a page, snapshot it, click, fill forms, take screenshots and record the screen
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| OpenTiny WebMCP SDK (this card) | no | 125 | ||
| Playwright MCP | yes | not required | 39,570 | drives the browser through accessibility snapshots; webmcp-cli works by DOM index and calls the page's own WebMCP tools |
| Chrome DevTools MCP | yes | not required | 55,126 | page debugging and profiling; this one controls the page and ships a WebMCP polyfill |
README in short
The toolkit consists of a WebMCP polyfill, client and server packages, a chat component and a CLI with skills for agents.