MCP Server Beta

The Archilogic MCP (Model Context Protocol) Server allows AI assistants and agents to access Archilogic resources directly through a standardized protocol.

WARNING

The MCP Server is currently in beta. Breaking changes may occur.

Overview

The Archilogic MCP Server is a remote MCP server that supports Streamable HTTP transport. It enables AI tools like Claude, ChatGPT, and Codex to interact with your Archilogic portfolio data, including buildings, floors, spaces, and metrics.

Available Operations

The MCP Server currently provides the following read-only operations:

Building & Floor Discovery

Operation	Description
list_buildings	Lists all buildings with their floors, addresses, and areas
list_floors	Lists floors with filtering options (by name, area, seat capacity, space count, or space properties)

Space & Metrics Analysis

Operation	Description
list_spaces_by_floor	Returns detailed space data grouped by floor. Includes area, category, subcategory, and assets
get_floor_space_metrics	Comprehensive area metrics broken down by space category (work, meet, care, socialize, circulate, support), seat capacity, and space counts
get_floor_element_metrics	Furniture/asset counts (workstations, lighting, seating, tables, storage) and building element data (windows, doors, walls)
get_space_asset_inventory_by_floor	Returns the assets (furniture, equipment) placed in each space, grouped by floor. Includes asset IDs, custom IDs, custom attributes, product details (name, category, subcategory), and asset counts per space

Floor Plan Intelligence

Operation	Description
get_floor_filter_options	Discovers all filterable attribute values on a floor (categories, subCategories, custom attributes)
export_floor_pdf	Generates a styled PDF floor plan with optional color-coded highlighting by category, subcategory, area, or custom attributes, and optional shape layers drawn on top of the plan. Supports byFilter rules for conditional styling
export_floor_svg	Generates an SVG floor plan of the default layout or a specific layout, with the same highlighting as export_floor_pdf, shape layers including icons, and options for visibility, units, and space labels

Both exports return a download link that is valid for 30 minutes.

Custom Operations via Extensions

You can dynamically add discoverable MCP operations by creating an Extension of type worker. Worker extensions are automatically exposed as operations for your organization by the MCP Server. Unlike the built-in operations, extension operations are not marked read-only. See how to create a worker extension to get started.

Setup

Endpoint

https://mcp.archilogic.com/mcp

The Archilogic MCP Server is a remote MCP server that can be used with any MCP client that supports remote MCP connections via Streamable HTTP. Archilogic hosts the server, so there is no API key to create and nothing to deploy. You sign in with your Archilogic account when your client first connects.

Claude Desktop

To use the Archilogic MCP Server with Claude Desktop, add the following configuration to your MCP settings:

{
  "mcpServers": {
    "archilogic": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.archilogic.com/mcp"]
    }
  }
}

This configuration uses mcp-remote to connect to the Archilogic remote MCP server via Streamable HTTP.

ChatGPT Desktop

To use the Archilogic MCP Server with the ChatGPT Desktop app, add it as a custom MCP server:

1. Open Settings → Integrations → Plugins.
2. Click Add → Add MCP Server and enter:
   • Name: archilogic
   • URL: https://mcp.archilogic.com/mcp
   • Type: Streamable HTTP
3. Click Save, then Authenticate to complete the OAuth flow. Make sure you are logged in to the Archilogic account you want to link before authorizing.

The Archilogic operations are now available in your ChatGPT Desktop conversations.

TIP

Adding custom MCP servers may not be available on all ChatGPT plans, and on Business and Enterprise workspaces an admin may need to enable it first. If the option is missing, refer to ChatGPT's help documentation.

Codex

On the same machine, the ChatGPT desktop app, the Codex CLI, and the Codex IDE extension share one configuration file, ~/.codex/config.toml, and one sign-in, so you only need to add the server once. If you already added it in the ChatGPT desktop app as described under ChatGPT Desktop, it is available in the Codex CLI and the IDE extension too. On a remote machine, for example over SSH, add it where Codex runs.

After adding the server, you sign in to Archilogic in your browser. If you use more than one Archilogic account, make sure your browser is signed in to the one you want to connect before you add the server.

Tested with

Codex CLI 0.157.0 and the ChatGPT desktop app 26.1002.52244 on macOS.

Codex CLI

Add the server with the Codex CLI, or add it to your configuration file:

Terminal

codex mcp add archilogic --url https://mcp.archilogic.com/mcp

~/.codex/config.toml

[mcp_servers.archilogic]
url = "https://mcp.archilogic.com/mcp"

codex mcp add saves the server and opens your browser to sign in. If your browser does not open, or you edited config.toml instead, run codex mcp login archilogic.

On a machine without a browser, for example over SSH, use the config.toml tab instead, then follow the steps for signing in on a remote machine.

Codex IDE Extension

In the Codex extension for VS Code, Cursor, or Windsurf:

• In the gear menu, open MCP servers
• Click Add server
• Enter archilogic as the name
• Select Streamable HTTP
• Enter https://mcp.archilogic.com/mcp as the URL
• Click Save
• Click Restart extension
• Click Authenticate next to archilogic to sign in

For more options, such as tool approval settings, see OpenAI's Codex MCP documentation.

Other MCP Clients

You can use the Archilogic MCP Server endpoint with any MCP client that supports remote Streamable HTTP connections and OAuth sign-in. Refer to your client's documentation for configuration instructions.

Authentication

The MCP Server authenticates users via Archilogic's OAuth 2.0 mechanism. When you first connect to the MCP server, you will be prompted to authorize access through Archilogic's OAuth flow. This grants the MCP server permission to access your portfolio data on your behalf.

MCP clients register themselves through OAuth dynamic client registration, so there is no client ID or secret to configure, and they refresh access tokens automatically.

Important

Make sure you are logged in to the Archilogic account you wish to link with the MCP connection before authorizing access. The MCP server will be connected to whichever account is currently active in your browser.

Codex Sign-In

1. Start the sign-in:
   • Codex CLI: codex mcp add starts it for you. If you edited config.toml, your browser did not open, or you need to sign in again, run codex mcp login archilogic.
   • ChatGPT desktop app and IDE extension: click Authenticate next to archilogic in the MCP server settings.
2. In your browser, sign in to Archilogic and approve access. If your browser is already signed in, check that it is the account you want to connect.
3. Check the connection:
   • Run codex mcp list in a terminal. The Auth column shows OAuth for archilogic once you are signed in, and Not logged in before.
   • Start a new Codex session so it picks up the new server. In the Codex CLI or the ChatGPT desktop app, type /mcp to see the connected servers.
   • Ask Codex to List all buildings in my portfolio.

You only need to sign in once per machine, because the CLI, the desktop app, and the IDE extension share the sign-in.

Signing in on a remote machine

On a machine without a browser, for example over SSH, add the server to config.toml (skip this if codex mcp add already added it) and run codex mcp login archilogic --no-browser. Codex prints a sign-in URL instead of opening a browser. Open it on your computer, sign in, and approve access. Your browser is then sent to a 127.0.0.1 address, which usually fails to load. Copy the full URL from the address bar, paste it at the Callback URL (input hidden) prompt, and press Enter. The pasted URL is not displayed, and the sign-in times out after 5 minutes.

Switch Accounts or Disconnect

These steps use the Codex CLI:

• Switch accounts: run codex mcp logout archilogic, sign out of Archilogic in your browser and sign in with the other account, then run codex mcp login archilogic.
• Disconnect: run codex mcp logout archilogic to delete the sign-in stored on your machine, then codex mcp remove archilogic to remove the server. Log out first, because removing the server keeps the stored sign-in.

Example Usage

Once configured, you can ask the MCP client to interact with your Archilogic data:

List all buildings in my portfolio

Show me the floors in building X

Get the metrics for floor Y

Which assets are placed in each space on floor Y?

Export floor Y as an SVG with meeting rooms highlighted

The AI assistant will use the appropriate MCP operations to fetch and display the requested information.

Extensions

The MCP Server's functionality can be extended by creating custom worker extensions. Worker extensions are automatically exposed as additional MCP operations for your organization, making them discoverable by any connected AI assistant.

To learn how to create and deploy a worker extension, see the Extension SDK guide.