Aceto
A local dev server with MCP interface for building HTML mockups together with an AI agent, enabling live collaborative editing through element selection and live preview.
README
Aceto
A local dev server with a browser overlay and MCP interface for building HTML mockups together with an AI agent.
Not a drawing tool, not a visual editor. Instead, a feedback loop between you and an agent:
- Tell the agent what you want
- The agent generates/modifies real HTML + Tailwind
- See the result live in your browser
- Point at an element and say "change this"
- The agent understands which element you mean and modifies it
Why Aceto
- Token-efficient: The agent doesn't need to read and parse the full HTML to understand what you mean. You point at an element in the browser, the agent gets a precise selector. Write tools use the current selection by default — one tool call, no roundtrip.
- Real HTML: No abstraction layer, no component model. The output is a plain HTML file you can read with
catand send by email. - Live feedback loop: DOM morphing keeps scroll position and selection state. You see changes instantly without page reload flicker.
Features
Hover Highlighting
User moves the mouse over any element — a blue outline appears with the element's tag and classes.

Element Selection
User clicks an element to select it (pink outline). The breadcrumb bar at the bottom shows the full DOM path. Scroll wheel navigates up/down the tree (parent/child).

Agent Highlight
The agent highlights elements in cyan to communicate visually — e.g. to ask "Do you mean this one?":
highlight_element("section#pricing > div:nth-of-type(2)", {
label: "Make this card wider?",
style: "pulse"
})
Both selections (pink = user, cyan = agent) can be visible simultaneously.

Inline Editing
User double-clicks any editable element to modify it directly in the browser:
- Text elements (headings, paragraphs, links, etc.) — edit the text content. Enter to save, Escape to cancel.
- Inputs — edit the value attribute. Enter to save, Escape to cancel.
- Textareas — edit the value. Ctrl+Enter to save, Escape to cancel, Enter adds a newline.
- Checkboxes / Radio buttons — double-click to enter edit mode, Space to toggle, Enter to save, Escape to cancel.

CSS Class Editor
User presses c on a selected element to edit its Tailwind classes inline. Enter to apply, Escape to cancel.

Table Controls
User selects a table cell — a floating toolbar appears with +Row, −Row, +Col, −Col controls.

DaisyUI Support
Add component libraries via the CLI — DaisyUI components work out of the box with Tailwind v4:
aceto add daisyui

More Features
- Content shortcuts — type
[]or[x]in a cell to insert a checkbox - Slash commands — type
/listin an empty element to open a modal for creating<ul>/<ol>lists. Edit badges appear on existing lists for quick re-editing via double-click. - Element defaults — define default classes for generated elements via
aceto.defaults.json - Paste images — Ctrl+V with selection inserts instantly; without selection, stages the image for agent-driven placement
- Asset picker — press
ato browse and reuse previously pasted images from the assets folder - Yank/Paste — press
yto copy a selected element,pto paste it after another selection - Screenshots — the agent captures full-page or element-level screenshots via
get_screenshot() - Live reload — DOM morphing keeps scroll position and selection state, no flickering
What Aceto Does Not Do
- No editor — you don't edit anything yourself, you steer the agent
- No build system — an HTML file you can read with
cat - No framework — Tailwind v4 via CDN, optionally DaisyUI/Flowbite
Keyboard Shortcuts (Select Mode)
| Key | Action |
|---|---|
| Click | Select element |
| Double-click | Inline edit (text, input value, checkbox toggle) |
| Scroll wheel | Navigate depth (parent/child) |
| Tab / Shift+Tab | Next/previous cell (during table editing) |
c |
Edit CSS classes of selected element |
y |
Yank (copy) selected element |
p |
Paste yanked element after selection |
u |
Undo |
r |
Redo |
d |
Insert empty div after selection |
a |
Open asset picker |
| Del | Delete selected element |
| Space | Toggle checkbox/radio (during edit mode) |
| Esc | Close modal / deselect / cancel edit |
e / Alt |
Toggle select/preview mode (clears selection) |
| Ctrl+V | Paste image |
Installation
bun install -g github:alexzeitler/aceto
Requires Bun runtime.
MCP Setup
Once aceto dev is running, add it as an MCP server to Claude Code:
claude mcp add aceto -s user --transport http http://localhost:3000/mcp
Getting Started
# Create a new project
mkdir my-mockup && cd my-mockup
aceto init
# Start the dev server
aceto dev
# Export for production
aceto export --production
CLI
aceto init # Create index.html + CLAUDE.md + aceto.md
aceto init --eject # Write default instructions into aceto.md
aceto init --preset daisyui # Create project with DaisyUI v5 pre-installed
aceto dev # Start dev server + MCP server
aceto dev about.html # Start with a specific page
aceto dev --port 3001 # Custom port
aceto new about # Create a new HTML page
aceto new dashboard/settings -l daisyui # New page with DaisyUI
aceto add daisyui # Add DaisyUI v5 to an existing project
aceto export # Export HTML with cleanup to dist/
aceto export --production # Export with Tailwind CSS build
Agent Instructions
On aceto init, a minimal CLAUDE.md is created that tells the agent to call get_instructions() via MCP. This tool returns the project's aceto.md if it has content, or the built-in default instructions.
To customize the instructions, run aceto init --eject to write the defaults into aceto.md, then edit to your needs.
Element Defaults
Create an aceto.defaults.json in your project root to define default classes for generated elements:
{
"checkbox": "h-4 w-4",
"img": "max-w-full h-auto rounded"
}
- checkbox — applied to checkboxes created via
[]/[x]shortcuts - img — applied to images inserted via paste (Ctrl+V)
The file is hot-reloaded — changes take effect immediately without restarting the server.
Tech Stack
Bun runtime, parse5 for HTML parsing, css-select for selectors, idiomorph for DOM morphing, MCP SDK for the agent interface. No Express, no React, no build step.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.