We Built an MCP Server for Leadpages. Here's What It Unlocks.

Leadpages is the only landing page platform with a remote MCP server. That's not a marketing claim. It's a fact about the category in 2026, and it changes what you can do with AI tools as part of your marketing workflow.

If you're not familiar with MCP, the short version: Model Context Protocol is an open standard from Anthropic that lets AI assistants (Claude, ChatGPT, others) take actions on your behalf in external systems. With MCP, you stop copy-pasting between Claude and your tools. You start asking Claude to do the work directly.

For Leadpages, that means a marketer running paid campaigns can build, edit, and publish landing pages from a Claude conversation. A developer can wire Leadpages page generation into a workflow. An agency can spin up 50 client pages programmatically. None of this required custom integration work or a separate API contract. Just an OAuth flow and a conversation.

Here's how the server actually works, what it can and can't do, and where this matters most.

What the MCP server does

The server lives at mcp.htmlpub.com. It exposes 47 tools across pages, sites, blogs, assets, and brand kits.

Page management. Create, edit, get, list, delete, and update pages. Set SEO metadata. Configure favicons. The full lifecycle of a published landing page, accessible from a chat window.

Site management. Create site collections. Move pages between sites. Update site settings. Useful for organizing pages by client, campaign, or project.

Blog and post management. Create blogs. Create, edit, and publish blog posts. Manage templates. The post you're reading right now was published through this server.

Asset management. Upload assets, import images from URLs (with SSRF protection so AI tools can't pull from internal IPs), browse the asset library. Plan-gated upload limits apply.

Brand kit management. Create and update brand kits. Extract brand identity from any URL. The AI analyzes a website and pulls colors, fonts, and voice, then applies them to new pages automatically.

URL analysis. Analyze external URLs for content extraction. Useful when migrating from another platform or building a page based on a competitor reference.

OAuth 2.0 with PKCE handles authentication. Six scopes (pages, sites, blogs, each with read and write) let you grant exactly the permissions an AI client needs. It's available on every paid plan starting at Grow ($99 a month).

What it doesn't do

The honest part. The MCP server can't:

• Set up or run A/B tests. Test creation lives in the page editor, not the API.
• Query analytics or pull conversion data. Reporting stays in the dashboard.
• Manage checkout or payment flows. We removed early claims about this. They weren't accurate.
• Configure personalization rules. Same as A/B tests. That's a builder workflow.
• Read or modify customer-facing user data.

If your use case requires any of those, you'll need the dashboard or a different tool. The MCP server is for content operations, not analytics or experimentation orchestration.

Use cases that actually work

Three places where this changes what's possible:

1. Programmatic landing pages at scale. Marketing teams running 20+ paid campaigns often need a unique landing page per ad group. Building these manually is the slow part. With MCP, you describe the variants in a Claude conversation, point it at your brand kit, and Claude creates the pages. The first time we tested this internally we shipped 15 pages in one hour, all branded, all published.

2. Agency operations. An agency running landing page work for 10 clients can wire MCP into their workflow. Pull a client URL, extract brand identity from their site, generate a landing page that matches the brand, push it to the client's domain. Closer to "no-code automation" than "custom integration."

3. Developer-led integration. If you've got a developer wiring Leadpages into a CMS, a CRM, or an internal tool, MCP gives them a clean interface. Authenticated, scoped, predictable. No bespoke API contract negotiation.

Compatible AI clients today: Claude (desktop, web, API), ChatGPT, Make. The list will grow as MCP adoption expands across the AI tool ecosystem.

Why we built this now

The honest answer: AI assistants are how marketers will increasingly do their work. Not in five years. Now. The team that figures out how to give the assistant useful tools for landing page work owns that workflow.

We built MCP first instead of waiting because waiting was the wrong move. By the time the rest of the category notices, the integration is already deployed, documented, and running in production for paying customers. That window matters.

The other reason: we already invested heavily in AI inside the page builder. Variant generation, brand kit extraction, plain-English summaries, AI page creation in 60 seconds. The MCP server makes those same primitives accessible from outside the editor. Same product, more surfaces to use it from.

Trade-offs to know

Cold start for AI clients. MCP is a new protocol. The list of tools that support it grows weekly, but if your team is on a tool that doesn't support MCP yet, the server doesn't help you directly. We'd recommend Claude or ChatGPT as the primary clients today.

Scope of operations. Content operations (create, edit, publish pages) are excellent through MCP. Analytics and experimentation are not. Don't try to use MCP as a full Leadpages replacement. It's a complementary surface, not the whole product.

OAuth setup. MCP authentication is OAuth 2.0 with PKCE. For most Claude users this is one-click ("authorize Leadpages"). For developers building custom integrations, there's setup work. Documentation lives on the MCP server feature page.

Plan gating. All MCP tools work on Grow ($99 a month) and above. Asset upload limits scale with plan. The free trial includes MCP access.

How to start

If you're a Leadpages customer on Grow or above, you can connect Claude or ChatGPT to the MCP server in about 60 seconds:

1. In Claude Desktop or ChatGPT, add a custom MCP connector pointing to mcp.htmlpub.com.
2. Authorize the connection through the OAuth flow.
3. Ask Claude or ChatGPT to create a landing page. It'll use your brand kit, your assets, and your existing pages as context.

If you're not on Leadpages yet, the free trial gets you the full product including MCP access for 7 days. Test it from a Claude conversation, build a page, see what's possible.

The category gap on this is real. We expect it to close eventually. If you want to be early on AI-powered landing page workflows, the time to test is now while the infrastructure is fresh.