UX Pilot

ai.uxpilotPublished by uxpilot.aiv3.0.0Updated Nov 4, 2026

Your AI designs live on the UX Pilot canvas or runs its design agent: reviews, React, video, media

The publisher's own server: QuickBooks Connector by Meridian

VerifiedStreamable HTTPWeb executableDeveloper ToolsMedia & DesignAI & ML

Overview

AI-generated overview

Connects an AI client to a UX Pilot workspace so the assistant can design, edit, review and export screens, React apps, videos and media.

What you can do
  • Find and read pages, screens, React projects, design systems, versions, comments and review scores
  • Edit screen HTML, React files and diagrams, and organize pages, folders, groups, notes and comments
  • Run UX Pilot's design agent for screens, flows, React apps and flowcharts, and import HTML, URLs, images or design systems
  • Review screens, HTML or live URLs with UX reviews, persona reviews, heatmaps and accessibility checks
  • Generate videos, images, voiceovers, music and sound effects, and export videos, preview links and React builds
  • Follow or stop running jobs and look up tiers, media models, costs and how-to guides
When to use it
Useful when you want an assistant to build or refine UI designs on a UX Pilot canvas, hand briefs to UX Pilot's design agent, run UX reviews with fixes, or produce marketing videos and media from the same workspace.
Requirements
Remote Streamable HTTP endpoint at the UX Pilot MCP server; a UX Pilot account with OAuth sign-in, or an API key sent as an Authorization Bearer header (for example via the UXPILOT_API_KEY variable in Codex); clients without Streamable HTTP support need Node.js to bridge through mcp-remote.
Before you install
Calls that spend credits, publish or delete ask for explicit confirmation with a token valid for 15 minutes; AI runs consume credits and some features need a paid plan; free accounts are limited to 50 writing or running actions a week; the API key grants access to your workspace, so keep it private and rotate it; delete actions always require approval.

Installation

In SourceWeft

  1. Open UX Pilot in the dashboard and add it to a workspace.
  2. Connect the required credentials. They are stored privately and never shown on public pages.
  3. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/ai.uxpilot%2Fuxpilot/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.

{
  "mcpServers": {
    "ux-pilot": {
      "type": "http",
      "url": "https://mcp.uxpilot.net/mcp"
    }
  }
}

README

UX Pilot MCP

Design, edit and review in UX Pilot from Claude, ChatGPT, Cursor, Codex, and other MCP clients.

Your AI builds screens live on the UX Pilot canvas, or hands the brief to UX Pilot's own design agent. Then it reviews, versions and ships the work: React prototypes, preview links, videos and media.

[MCP] [Website] [MCP Badge]


The UX Pilot MCP server connects any MCP-compatible AI client to your UX Pilot workspace. Version 3.0 has 15 general tools. Each one covers a kind of work (find, read, edit, review, export and so on), and you choose the action inside the tool. That keeps the tool list short and leaves more of your AI's context for the work.

The server is hosted by UX Pilot. This repository documents how to connect to it and what it exposes. There is no server code here to build or run.

  • Server URL: https://mcp.uxpilot.net/mcp
  • Transport: Streamable HTTP (POST)
  • Protocol: MCP 2025-11-25 (JSON-RPC 2.0)
  • Auth: Native OAuth sign-in (recommended), or Authorization: Bearer API key fallback
  • Tools: 15 (version 3.0)
  • Free plan: connect with a free UX Pilot account and get 50 actions a week

What your AI can do

  • Design on the canvas, live. Your AI places a screen as HTML, then builds it one section at a time while you watch it appear on the open canvas. Each step comes back with a picture of the result, and design checks point out problems with fixes.
  • Run UX Pilot's design agent. Hand a brief to UX Pilot's own agent for screens, whole flows or a React app, on the standard, pro or ultra tier. Or mix the two: the agent drafts and your AI refines.
  • Review. UX reviews, persona reviews, attention heatmaps and gaze flow, on a UX Pilot screen, raw HTML or a live URL. Heatmaps and gaze flow also take a screenshot. Fixes land as a new version.
  • Make videos and media. Product reels, ads, explainers, social clips and logo animations, exported as MP4, WebM or GIF. Images, image edits, voiceovers, music and sound effects.
  • Use your design system. Import one from code, a repository, a website or tokens, then design on brand with its guide.
  • Ship. Turn screens into a React app, edit its files, publish it, or share a public preview link.
  • Organize. Pages, folders, groups, versions, canvas notes, diagrams and comments.

Free plan

Free UX Pilot accounts can connect and use every tool:

  • 50 actions a week that write or run something (imports, edits, organizing, designs, reviews, media, videos). Reading, searching, checking jobs and stopping runs are not counted. The week resets Monday 00:00 UTC.
  • Design runs use the standard tier. Some features need a paid plan, the same as in the app. When one does, the tool says what is needed and links to the plans.
  • AI runs spend credits, as they do in the app.

Paid plans have no weekly action limit. See uxpilot.ai/plans.

Authentication

Native sign-in (recommended)

If your AI tool shows a UX Pilot sign-in or authorization screen, choose that flow. You log in with your UX Pilot account, review the requested access, and approve the connection. No API key to copy or paste, and access stays tied to your account and plan.

API key fallback

For clients that only support static headers, generate a key from the API Keys tab in your UX Pilot workspace and send it as a bearer token:

Authorization: Bearer ep_your_api_key

Enterprise workspaces: your admin manages MCP/API keys. Keep the key private and rotate it if access changes.

Setup

Prefer native sign-in wherever the client supports it. Use the API key config as a fallback. Replace ep_your_api_key with your real key.

Claude

Native connector (recommended):

  1. Open Claude settings, go to Connectors, and pick UX Pilot from the directory. Or add a custom connector with the server URL: https://mcp.uxpilot.net/mcp
  2. Sign in with your UX Pilot account and approve access.
  3. In a conversation, ask Claude to find your UX Pilot pages, or to design something on the canvas.

Claude Code:

bash
claude mcp add --transport http uxpilot https://mcp.uxpilot.net/mcp

Then run /mcp in Claude Code and sign in.

API key fallback (Claude Desktop, via mcp-remote): edit claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json, Windows: %APPDATA%\Claude\claude_desktop_config.json)

json
{  "mcpServers": {    "uxpilot": {      "command": "npx",      "args": [        "-y",        "mcp-remote",        "https://mcp.uxpilot.net/mcp",        "--transport",        "http-only",        "--header",        "Authorization:${AUTH_HEADER}"      ],      "env": {        "AUTH_HEADER": "Bearer ep_your_api_key"      }    }  }}

Requires Node.js. mcp-remote bridges stdio to Streamable HTTP.

ChatGPT

Use ChatGPT's custom connector / MCP support (Developer mode, availability depends on your plan):

  1. Open ChatGPT settings and add a custom connector / MCP server.
  2. Paste the server URL: https://mcp.uxpilot.net/mcp
  3. Complete the UX Pilot sign-in when prompted.
  4. Enable the connector in a chat and ask ChatGPT to find your UX Pilot pages.
Cursor

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (project). For native sign-in, add only the url. Keep the headers block only for the API key fallback.

json
{  "mcpServers": {    "uxpilot": {      "url": "https://mcp.uxpilot.net/mcp",      "headers": {        "Authorization": "Bearer ep_your_api_key"      }    }  }}
VS Code + Copilot

Add to .vscode/mcp.json:

json
{  "servers": {    "uxpilot": {      "type": "http",      "url": "https://mcp.uxpilot.net/mcp",      "headers": {        "Authorization": "Bearer ep_your_api_key"      }    }  }}
Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

json
{  "mcpServers": {    "uxpilot": {      "serverUrl": "https://mcp.uxpilot.net/mcp",      "headers": {        "Authorization": "Bearer ep_your_api_key"      }    }  }}
Codex

Add to ~/.codex/config.toml:

toml
[mcp_servers.uxpilot]url = "https://mcp.uxpilot.net/mcp"bearer_token_env_var = "UXPILOT_API_KEY"

Then export the raw key (no Bearer prefix; Codex adds it):

bash
export UXPILOT_API_KEY="ep_your_api_key"
Any MCP client

Clients that do not speak Streamable HTTP can bridge through mcp-remote:

json
{  "mcpServers": {    "uxpilot": {      "command": "npx",      "args": [        "-y",        "mcp-remote",        "https://mcp.uxpilot.net/mcp",        "--transport",        "http-only",        "--header",        "Authorization:${AUTH_HEADER}"      ],      "env": {        "AUTH_HEADER": "Bearer ep_your_api_key"      }    }  }}

Tools

15 tools. Most take a kind or action that picks what to do, and a params object with that action's fields. Ask the server with guide({ topic: "<tool>.<action>" }), for example guide({ topic: "edit.patch" }), for an action's fields and an example.

ToolWhat it doeskind / actionNeeds your OK
findFind work: pages, folders, screens and React apps by keyword, themes, symbols, design systems, diagrams, React files and codepages, workstreams, designs, react_projects, themes, symbols, design_systems, diagrams, react_files, react_code
readRead one thing in full: a page with its canvas, layout or picture, a screen and its HTML, versions and diffs, a preview link, design checks, a design system and its guide, review scores, commentspage, design, designs, versions, version_diff, preview, checks, theme, symbol, design_system, design_system_guide, react_file, review_scores, comments
catalogWhat you can pick from: design tiers, media models, review and export types, and what each paid call coststiers, media_models, media_kinds, reviews, exports, credits
guideShort how-to docs: ways to design, the canvas, small steps, HTML, videos, React, design systems, reviews, the free plan, confirmationstopic
get_jobFollow a running job: design runs, React edits, videos, imports, reviews, video exports, design system importsagent, generation, review, export, design_system_import
stop_jobStop a design, React or video run
editChange content with no AI and no credits: patch a screen's HTML one section at a time, replace it, save a version, make a version from comments, edit React files or a diagrampatch, replace, save_version, version_from_comments, react_files, diagram
manageOrganize, never delete: pages, templates, bookmarks, folders, groups, screens, canvas notes and shapes, images, diagrams, comments and replies, accessibility reportscreate_page, page_from_template, update_page, bookmark_page, move_page, create_workstream, update_workstream, create_group, update_group, save_group, move_group, duplicate_design, move_design, update_design, save_design, add_items, update_items, update_assets, create_diagram, comment, reply, update_comment, update_reply, resolve_comment, resolve_comments, accessibility_report
importBring things in: your HTML as a new screen (free), a live URL, images by URL or upload, a design system from code, a repository, a website or tokenshtml, url, assets, upload, design_systemCredits (url, html with AI refine, or a design system from code, a repository or a website; tokens are free)
deleteDelete pages, folders, groups, canvas items, images, diagrams, comments and replies, or archive a screenpage, workstream, group, design, canvas_items, canvas_assets, diagram, comment, replyAlways
designUX Pilot's AI: the design agent for screens, flows or a new React app on the standard, pro or ultra tier, screens into a React app, React changes from a prompt, flowchartsagent, react, react_edit, flowchart, flowchart_editCredits (flowcharts are free)
make_videoMake or change a video from a brief: product reel, ad, explainer, social clip or logo animation, in 16:9, 9:16 or 1:1Credits
generate_mediaMake media for a page: images, image edits, image processing, voiceovers, music, sound effectsimage, edit_image, process_image, voiceover, music, sound_effectCredits
reviewReview a screen, raw HTML or a live URL: UX review, persona review, attention heatmap, gaze flow (these two also take a screenshot). apply: true fixes ux, persona or accessibility issues as a new versionux, persona, heatmap, gaze_flow, accessibilityCredits (a persona review is free)
exportExport a video as MP4, WebM or GIF, publish a public preview link, publish or rebuild a React appvideo, preview, prototype, react_buildCredits for a preview link or a React app publish; a rebuild asks first; video export is free

Confirmation

Calls that spend credits, publish or delete stop and ask first:

  1. The first call answers with a summary and a confirmationToken instead of running.
  2. Your AI tells you what will happen and asks.
  3. If you agree, it calls again with the same params plus confirmed: true and the token. The token is tied to that exact call and expires after 15 minutes.

Upgrading from 2.x

Version 2.x listed 87 separate tools, such as search_designs, start_design_agent and patch_design_html. Nearly all of them are now actions of a 3.0 tool, and guide({ topic: "names" }) maps every old name to its new call. For example, search_designs is find with kind designs, start_design_agent is design with action agent, and patch_design_html is edit with action patch. The old tool list is still served at https://mcp.uxpilot.net/mcp?toolset=legacy for now, and it will be removed in a later release.

Quick start

  1. Pick the AI client you want to use with UX Pilot.
  2. Add the server URL with the native sign-in steps above.
  3. If your client needs static headers, generate an API key and replace ep_your_api_key in the fallback config.
  4. Ask your assistant, for example:
    • "Design a pricing page for my app on the UX Pilot canvas, one section at a time."
    • "Have UX Pilot's design agent make a 3-screen onboarding flow on the pro tier."
    • "Run a UX review on my checkout screen in UX Pilot and fix what it finds."
    • "Make a 15-second 9:16 product reel for my app and export it as MP4."

Support

Source: README.md at commit 676a88b

Tools

0
This server asks you to sign in, so SourceWeft could not read its tools without an account and has not indexed them. What a connected account sees can depend on that account and its permissions.

Version history

2
  1. v3.0.0LatestNov 4, 2026
  2. v1.0.0Oct 22, 2026

Score

75% 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 daysRequired25 pts
  • Auth declaredRequired10 pts
  • 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

    Declare a license in the repository or the registry entry.

Other Pilot integrations