Using Brandize with ChatGPT and Other MCP Clients

Where supported, ChatGPT, Cursor, and Codex can call the same tools on Brandize's MCP server. Each client enables remote MCP connections differently. In ChatGPT, access to a custom connection can depend on your account and workspace, so check that first.

One endpoint, different client experiences

Brandize's remote endpoint is https://brandize.me/api/mcp. It uses Streamable HTTP and currently needs no Brandize sign-in for its free tools. The tool list includes palette and site-markup generators; generate_logo returns a watermarked preview and a checkout link for a selected package.

If you mainly use Claude, the Claude walkthrough follows a brief from the first palette call through the logo preview.

Your client determines how you add the endpoint and how it displays a tool result. It does not change the arguments Brandize needs. A palette call still takes a base hex color. A logo call still needs a real design brief and a package choice. If you move to another conversation or client, save the job token for a later get_logo_result call.

ChatGPT: check access before the branding brief

OpenAI's ChatGPT MCP connection guide describes adding a public MCP endpoint through a developer-mode plugin. It also says developer-mode availability depends on account and workspace policy. If that option is present for you, use the Brandize endpoint details when creating the connection, then inspect the tools ChatGPT discovers. An installed plugin may also provide MCP tools in a workspace that supports plugins, as the OpenAI MCP documentation explains.

Start with a free request: "Use Brandize to make a five-color palette from #3B82F6, and show me the hex values." If ChatGPT only writes a palette in prose, check whether the MCP connection is enabled in that conversation and whether it actually selected generate_color_palette. A convincing answer in chat is not evidence that a tool ran.

Before requesting a logo preview in ChatGPT, ask which brief details are still missing: the package, style, layout, color, industry, and the surface where the logo needs to read well. Choose those details yourself.

Cursor and Codex: keep the assets near the site

Cursor supports remote MCP servers over Streamable HTTP, according to its MCP documentation. Codex also supports remote MCP servers; OpenAI's Codex MCP guide covers its connection and tool views. Brandize has client-specific setup notes for both, including the URL and plugin options. Follow those current instructions rather than copying an old configuration from a blog post.

These clients are useful when the next step is in a codebase. Put the chosen palette colors into CSS tokens. Compare returned meta tags with the actual page title, description, and canonical URL. Check that Organization schema names the right business and site. The MCP tools return text and generated assets; they do not silently edit or deploy your site.

A placeholder URL can point schema at the wrong entity, and a neat meta description can describe a page you haven't built. Review both against the actual page before you use them.

What happens when you ask for a logo

Whichever client you use, the paid path starts with your brief. You choose the package, then provide the brand name, what the business does, industry, style, layout, main color, and primary use surface. The agent should ask for missing choices before calling generate_logo. You can use get_pricing_tiers to see the current package options first.

The tool returns a low-resolution watermarked preview and a checkout URL. Review the preview. If you want the full-resolution, watermark-free deliverable, open the checkout link and complete payment there. Brandize then makes the purchased files available through get_logo_result after payment settles. Read the payment and consent details before starting that step.

The free tools are enough if you're exploring colors or page markup. If a connection fails, use the troubleshooting section. For the endpoint and current client instructions, use the Brandize MCP hub.