What is the BranderUX MCP?

A remote MCP server at mcp.branderux.com/mcp. It gives your agent tools to build and control full BranderUX applications: projects, brand, custom elements, screens and API keys, plus the hosted lane, business data entities with live store or API feeds, the agent itself with its skills and policies, a designed home screen, and one-call publish of the website and MCP app. Reference docs for the SDK ride along.

Do I need an API key?

No. The first tool call opens your browser for a one-click BranderUX sign-in (OAuth 2.1 with PKCE). Tokens are scoped and revocable; nothing is pasted into your agent config.

Who can connect it today?

BranderUX beta accounts. Everyone else lands on the waiting list, we are opening the beta deliberately, in small cohorts.

How do I connect BranderUX to Claude?

Three steps: (1) copy the connector URL https://mcp.branderux.com/mcp, (2) in Claude go to Settings → Connectors → Add custom connector, name it BranderUX and paste the URL, (3) connect and sign in when the browser opens. Then ask Claude to build you a project.

Is the BranderUX MCP in the official MCP Registry?

Yes, published as com.branderux/mcp in the official MCP Registry (registry.modelcontextprotocol.io), with the remote endpoint https://mcp.branderux.com/mcp (Streamable HTTP). Clients and directories that sync the registry can discover it automatically.

Which clients are supported?

Any MCP client that speaks Streamable HTTP with OAuth 2.1: Claude Code, Claude web and desktop connectors, ChatGPT, Cursor, VS Code (Copilot agent mode), Windsurf, Zed, Cline, Gemini CLI and others. The connect tabs on this page give the exact config for each.

Is this the same as putting BranderUX in my own product?

No, they are two different jobs. This MCP is how you and your coding agent BUILD on BranderUX. To SHIP branded UI to your own users, your product's agent uses @brander/sdk (any framework: LangGraph, CrewAI, Mastra, Pydantic AI, Google ADK, provider SDKs, or a custom loop), or @brander/mcp-tools if your product is itself an MCP server.

How is authorization secured?

OAuth 2.1 with PKCE, dynamic client registration and Client ID Metadata Documents, plus RFC 8707 resource indicators: every token is audience-bound to the server it was issued for. The MCP gateway never forwards your token downstream, it exchanges it (RFC 8693) for a short-lived API token, so no service can replay a credential meant for another.

What can the agent actually do?

Create projects, set brand identity, author and publish custom elements (your agent writes the component code, we validate and publish it), compose screens, and mint the embed API key. Elements and screens render as live interactive panels right in the chat (MCP apps), and it serves verified integration snippets so your SDK wiring is right first try.

Can it delete things?

Destructive tools exist but every one requires an explicit confirmation flag, so an agent cannot delete a project or element without being told to.

What permissions does it get?

Scoped: account:read, projects:read/write, elements:read/write and keys:manage. You see the list on the consent screen and can revoke access at any time.

BranderUX MCP for any agent

Connect the BranderUX MCP server and build full agentic applications from your prompts, in any AI tool: the project, the brand, the data, the screens, the agent itself, and the publish. Everything real and live, no API key to paste.

1

Copy the BranderUX connector URL

You'll paste this into Claude in the next step.

https://mcp.branderux.com/mcp

2

Go to Claude → Settings → Connectors

In Claude desktop or claude.ai, add a custom connector, name it BranderUX and paste the URL.

Open Claude Connectors ↗

3

Connect, sign in and start

Sign in when the browser opens, then ask Claude to build you a project.

Beta accounts today, request access

One prompt in, a live app out

✳ Claude

Reply to Claude…

Claude builds it, through the BranderUX MCP

Loading Atelier Nova…

The result, live right now: try it in the frame or open it ↗

Not only full apps

The same conversation handles the smaller jobs, a single element, or wiring the SDK into your own app. Real tool runs, not mockups.

You

Create a custom element for my project: an interactive product grid.

Agent · calling BranderUX

I'll read the authoring contract, write the component TSX myself, and publish it, validation runs before anything ships.

✓read_doccustom-elements-contract

✓create_elementproduct-grid · compile + allowlist checks passed

✓list_elementsproduct-grid · published v1

✓put_screencatalog · version pinned

Result

The published custom element rendering in a screen

Sandboxed, versioned, and available to the runtime AI immediately.

Build with BranderUX skills in your agent

Every screen below was generated at runtime. Recreate runs the same prompt in Claude.

Use case

Full agentic application

“Using the BranderUX MCP, build the whole application for my shop: create the project, set the brand, connect my data, configure the agent, design the home screen, and publish it.”

Recreate ↗Learn more

Skill

Interactive element

“Using the BranderUX MCP, create a custom element for my project: an interactive product grid.”

Recreate ↗Learn more

Use case

BranderUX in your app

“Using the BranderUX MCP, integrate BranderUX into my Next.js app: create a project, mint an API key, and wire the SDK with a streaming backend route.”

Recreate ↗Learn more

Every tool, inside your agent

42 tools and the docs to use them, your agent reads the reference before it writes, so element code and screen layouts are right the first time. Every destructive tool requires an explicit confirm: nothing is deleted without your yes.

Projects

7 tools

Create and brand real projects

whoami · list_projects · get_project · create_project · update_brand_settings · update_project_settings · delete_project

Custom elements

7 tools

Your agent writes the TSX; we validate, publish and render it live in the panel

list_elements · get_element · list_element_versions · create_element · publish_element_version · preview_element · delete_element

Screens

4 tools

Compose the pattern library the runtime AI learns from

list_screens · get_screen · put_screen · delete_screen

API keys

4 tools

Mint origin-locked embed keys

create_api_key · list_api_keys · set_key_origins · revoke_api_key

Docs & snippets

4 tools

The reference, readable by the agent itself

get_started · read_doc · search_docs · get_integration_snippet

Hosted agent

2 tools

Give the project its own agent: persona, policies, answer quality, budgets

upsert_agent_config · get_agent_config

Business data

5 tools

Managed records or live store/API feeds; every entity becomes a query tool for the agent

define_entity · list_entities · seed_records · update_record · list_entity_records

Connectors

2 tools

Vaulted credentials, and a probe so tools are wired against the real API shape

set_connector_credential · probe_api

Skills

3 tools

SKILL.md behavior packs that ride every conversation

upsert_skill · list_skills · delete_skill

Home screen

1 tools

A designed first paint with live data bindings, zero model cost per visit

set_home_screen

Publish

2 tools

One call and the full application is live: a website for people, an MCP app for agents

publish_site · get_site

Playground

1 tools

Render a branded screen as a live MCP app in the chat, with or without a project

generate_screen

Questions

Connect your agent

Beta accounts connect in under a minute, everyone else can hold a spot on the waiting list.

Set up the connector ↑Request access

BranderUX MCP for .css-1qggthf{color:var(--accent);}any agent

One prompt in, a live app out

Not only full apps

Build with BranderUX skills in your agent

Every tool, inside your agent

Questions

Connect your agent

BranderUX MCP for any agent