MCP Integration

Connect your AI tools to Architect using the Model Context Protocol (MCP)

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants — like Claude, Cursor, ChatGPT, and other tools — interact directly with external services through a secure, structured interface.

Architect exposes a hosted MCP server that lets your AI tools manage pages, audiences, prospects, analytics, and more — without needing to run anything locally.

What can you do with Architect’s MCP?

📄 Pages

Create, read, update, delete, archive, and restore pages — the core content units in Architect. Pages can be of various types (inbound, outbound, call follow-up, sales deck, custom) and support actions (open URL, submit a form, copy text), widgets (logo marquees, videos), scripts, footer links, and rich SEO metadata.

👥 Audiences

For each page, you can manage audiences — specific versions of a page targeted at different segments. You can create, update, archive, restore, and search audiences. Each audience can have overrides for actions, URLs, metadata, and supplementary context.

📝 Audience Versions

Audiences are versioned. You can list, get, create, publish, cancel, and manage the lifecycle of audience versions — moving them through draft → queued → processing → published states.

💬 Audience Messages (AI Editing)

You can send messages to an audience’s draft editor (an AI-powered interface), retrieve conversation history, check on active AI runs, and interrupt them mid-stream. This enables programmatic AI-assisted content editing.

🔗 Page Links

Create and manage shareable URLs for pages and audiences — mapping custom paths to pages or audiences, with optional routing tables for A/B-style traffic splitting across multiple targets.

👤 Persons

Full CRUD on person records (prospects) — create, list, search, update, delete, and restore. Fields include name, email, domain, phone number, and LinkedIn profile URL.

🏢 Accounts

Manage company/account records by domain — create, list, search, update, and retrieve accounts along with their linked persons and activity history.

📋 Activity Logs

Ingest structured activity events via person logs and account logs — useful for syncing CRM events, form submissions, webhook data, or any custom activity into Architect from external systems.

📊 Analytics

Query analytics data including page summary stats (bounce rate, conversion rate, session time, scroll depth), per-audience breakdowns, individual audience session details, and interaction histories for sessions.

🔔 Webhook Subscriptions

Subscribe to real-time events from Architect — including page.version.published, form.submitted, form.enriched, and form.enrichment_failed. You can create, list, update, and delete subscriptions.

How do you use all of this?

Architect’s MCP server exposes two tools:

Tool	What it does
architect_docs	Read Architect MCP documentation and targeted Customer API reference slices. Use this first to discover available endpoints, learn the request/response shapes, and understand how resources relate to each other.
architect_api	Call any Customer API endpoint for the authenticated tenant. Once you know the method, path, and payload from the docs, use this tool to execute the actual request.

The recommended workflow:

1. Ask your AI tool to use architect_docs to look up what you want to do.
2. Then ask it to use architect_api to execute the call.

For example:

• “Use architect_docs to find out how to create a page, then create one for my new campaign.”
• “Look up the analytics endpoints in architect_docs, then pull my page stats for this month.”
• “Check architect_docs for how person records work, then create a prospect for [e-mail].”

How authentication works

Architect’s MCP uses OAuth 2.0. When you connect your AI tool to Architect’s MCP server, you’ll be redirected to sign in with your Architect account. The MCP server automatically scopes all requests to the tenant tied to the account you authorize.

Available scopes

Scope	Description
api:read	Required for documentation reads and GET requests through architect_api.
api:write	Required for POST, PATCH, and DELETE requests that change Customer API state.

Getting started

1. Create a new remote MCP server in your client

   Look for an MCP, integrations, or developer tools settings screen in your AI tool.

2. Paste the server URL

   Architect uses a hosted Streamable HTTP MCP endpoint, so you do not need to run anything locally.

   https://mcp.tryarchitect.com/mcp
   
3. Sign in with Architect when prompted

   The MCP server automatically scopes requests to the tenant tied to the account you authorize.

4. Start with architect_docs

   Ask your AI tool: “Use architect_docs to show me what I can do with the Architect API.”

Client-specific setup guides

Claude Desktop

Step-by-step setup for Anthropic’s Claude Desktop app.

Cursor

Step-by-step setup for the Cursor AI code editor.

Advanced setup

Most clients discover OAuth metadata automatically. Use these only if your client asks for them.

URL	Purpose
https://mcp.tryarchitect.com/.well-known/oauth-protected-resource	Advertises the MCP resource server and supported scopes for OAuth-aware clients.

Some tools (like Claude Desktop) don’t support remote URLs directly and require a bridge like mcp-remote. See the individual setup guides for exact instructions.