AI agents

MCP Server

AI agents

MCP Server

Connect Claude, ChatGPT, Gemini and other agents to the Contit MCP servers with your account, or with a Personal Access Token.

Contit and its apps expose remote MCP servers (Model Context Protocol, Streamable HTTP): agents like Claude, ChatGPT, Gemini or Claude Code use your workspaces' data as native tools, with your permissions.

Server Address What it does
Contit https://api.contit.cloud/mcp Workspaces, content types and contents, read-only
Autopilot https://autopilotapi.contit.cloud/mcp Contacts, lists and segments, campaigns and sends, templates, automations and tickets
Shop https://shopapi.contit.cloud/mcp Catalog and translations, orders, customers and sales statistics
Site https://site.contit.cloud/mcp Pages, paragraphs, layouts, collections, assets and site diagnostics
Folder https://folderapi.contit.cloud/mcp Folders, files and users of the reserved area, with who sees what

The addresses, with a copy button, are also in the Contit profile → AI agents.

Connect an agent with your account

This is the recommended way: no token to copy. The agent starts from the server address, lets you sign in with Contit and asks for your consent.

Claude

On claude.ai or Claude Desktop: Settings → Connectors → Add custom connector, paste the address and click Connect. The Contit sign-in opens: sign in with your account and click Allow. The connector also works from the Claude app on your phone.

ChatGPT

On chatgpt.com add a custom MCP server with the address and OAuth authentication, then sign in with Contit and allow. ChatGPT custom MCP servers currently work from the web.

Gemini

In Gemini open Connectors and connect an MCP server: paste the address, leave Client ID and secret empty, then sign in with Contit and allow.

Claude Code

bash
claude mcp add --transport http contit https://api.contit.cloud/mcp

Then open /mcp in Claude Code and pick the server to sign in with Contit.

Workspaces and permissions

The connection covers all the workspaces you can access: the agent lists them with the list_workspaces tool (contit_list_workspaces on the Contit server) and picks the workspace in the conversation, passing workspaceId to the tools. Each time the server checks that you can access that workspace. The agent asks for confirmation before changing anything.

See and disconnect agents

In the Contit profile → AI agents you find the connected agents, with the service and last use. Disconnect removes access: the agent loses it within 15 minutes and must ask for your consent again to get it back.

Each server is connected separately and asks for its own consent: connect only the ones you use.

Connect with a Personal Access Token

For Codex, scripts or manually configured clients you can use a Personal Access Token (PAT), sent as a Bearer.

Kind Where to create Scope
Per-workspace Console → workspace Settings → Developer → Personal Access Tokens a single workspace, with per-content-type permissions
Admin (multi-workspace) Contit profile → Access Token all workspaces you administer; restricted to subscription Owner/Manager

The token (format ctpat_...) is shown only once: copy it immediately. With an admin token you use a single configuration for all your workspaces, and the agent picks the workspace as with the account connection.

For workspaces of a reserved app that exposes its own MCP server, the panel shows the app's MCP server address; for Contit use the Contit API /mcp endpoint.

Claude Code

bash
claude mcp add --transport http contit https://api.contit.cloud/mcp \
  --header "Authorization: Bearer ctpat_xxxxxxxx"

Claude Desktop with mcp-remote

In claude_desktop_config.json, with the X-Api-Key header (avoids issues with the space in Authorization: Bearer):

json
{
  "mcpServers": {
    "contit": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.contit.cloud/mcp",
        "--header",
        "X-Api-Key: ctpat_xxxxxxxx"
      ]
    }
  }
}

The PAT (ctpat_…) is accepted equivalently as Authorization: Bearer, an X-Api-Key header, or a ?api_key= query parameter.

Contit server tools

Tool Description
contit_list_workspaces Lists accessible workspaces
contit_list_content_types Lists the workspace's content types (schemas)
contit_query_content Searches published contents of a content type, with filters, sorting and pagination
contit_get_content Retrieves a single published content by content type and id

All tools require read permission. With a per-workspace token the workspace is implicit; with an admin token or the account connection the tools accept the workspaceId parameter, and access to that workspace is verified on every call.

If the PAT has per-content-type permissions, the tools only see and return the content types it can read, like the REST APIs.

Centralized validation for reserved apps

Reserved apps can validate the same Personal Access Token centrally via the /connect/introspect endpoint, so the authority that knows the tokens stays single (the IDP). In the .NET SDK:

csharp
var validator = new ContitTokenValidator(appClientId, appClientSecret);
var info = await validator.ValidateAsync(incomingToken);
if (info.Active)
{
    // info.UserId, info.WorkspaceId, info.CanRead, info.CanWrite
}

To call the APIs on behalf of the user with a PAT:

csharp
var client = DelegatedContitClient.FromPersonalAccessToken("ctpat_xxxxxxxx");