Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@html-mcp-web inspect and process the open comments on the slides"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Review an AI agent's HTML slides or report from the rendered page, on a desktop, tablet, or phone, while Claude Code or Codex edits the source. Hand the finished slides over as an editable PowerPoint deck.
⭐ If this helps your reviews, please give it a star. It helps others find the project.
Paste this into Claude Code or Codex:
Install html-mcp-web by following https://raw.githubusercontent.com/MiiKiyoshi/html-mcp-web/main/INSTALL.mdThe agent checks for Python and Firefox, shows you what it will install, and installs once you agree. Start the agent again afterwards.
Related MCP server: Terrarium
Once per folder, start the agent in the folder that holds (or will hold) your artifact and say:
do html initChoose slides (16:9) or a report (A4). Slides default to neutral-slides, a plain template
without organization marks, and neutral, writing rules the agent follows for slides. Reports
default to neutral-report, with a cover and one page per section, and no guideline. The agent
shows the choices and the document file it will create or reuse. Accept the defaults or name
what to change.
To review, say:
do html listenThe agent gives you the review page address, such as http://localhost:8765, and gets ready
to handle your comments when you press Call agent. Open that address in your browser. Say
do html listen again after the agent restarts.
Select text on the rendered page and press the Comment button beside it. + Note comments on the whole artifact.
Press Call agent when your comments are ready. The agent edits the source and replies in the same thread.
Resolve a thread whose edit satisfies you, or Reply in it when it does not. Archive sets a thread aside, and it still takes replies.
The thread shows the source the agent would change as a −/+ pair, and the Source view marks the same text in red. Apply suggestion writes the change into the file. Proposing again replaces the proposal.
Press PPTX in the topbar for an editable PowerPoint deck of the slides, or PDF to print every page.
Use the port number in the review page address the agent gives you. Forward that port over SSH from Termux on Android or iSH on iOS, then open the same address in the phone's browser.
ssh -L <port>:localhost:<port> <server>Edit .html-mcp-web.yaml in the project folder. A saved change applies at once, except
port, which applies when Claude Code or Codex restarts.
artifacts:
slides:
label: Slides
layout: slides
main: slides.html
watch:
- '*.html'
- '*.css'
- '*.js'
- '*.svg'
- '*.png'
- '*.jpg'
- '*.jpeg'
- '*.gif'
- '*.webp'
ignore: []
port: 8765Field | Effect |
| Name the review page shows for the artifact. |
|
|
| The HTML file the page shows. |
| A template name and the content file it builds |
| A writing guideline the agent follows, such as the built-in |
| Files whose saves refresh the page. |
| Patterns excluded even when they match |
| This folder's review page port. Any free port works. Init picks the first free one from 8765, so each folder can have its own. |
Put a template in ~/.config/html-mcp-web/templates/<name>/. Start from a copy of
templates/neutral-slides, as templates/SKINS.md explains. Put a
writing guideline for the agent in ~/.config/html-mcp-web/guidelines/<name>/GUIDELINE.md.
To build on the built-in neutral guideline instead of
copying it, open your file with this front matter, write only the rules you add or change,
and say that yours win where the two differ:
---
extends: neutral
---A name you use replaces the built-in one of the same name. do html init offers both.
MIT licensed. See LICENSE. The source editor uses Ace under the BSD license, and its license is included with the bundled files. Full-screen wheel navigation adapts the intent-detection strategy from Swiper's Mousewheel module by Vladimir Kharlampidi and the Swiper contributors, under the MIT license.
This server cannot be deployed
Human feedback for AI agents: share HTML, get a live review link, read anchored notes as markdown.
Human visual review for agent-built HTML: publish artifacts, pull anchored comments, push fixes.
Comment on AI-generated webpages; feedback flows back to your coding agent. Free, MIT, local-first.
MarkupBase turns AI-generated Markdown and HTML into durable, versioned artifacts that people can review and discuss. Its MCP server lets agents publish new versions, preserve contextual comments, include hosted images, and respond to feedback through secure account-linked identities, creating a clear human review boundary without requiring real-time editing.
