Connect your AI over MCP

Connect Claude, ChatGPT, Codex, or Cursor to Liteo, sign in once, and let it search, read, and update your workspace.

Reviewed September 23, 2026

Browse all guides
On this page

Liteo has an MCP server. Add it to your AI client, sign in to Liteo, and your AI can work in your workspace through named tools.

The server address

Liteo's MCP server lives at one address:

https://app.liteo.io/api/mcp

Once connected, it can search your contacts, businesses, and deals. It can read a customer's sites and installed products. It can draft a quote, record a service visit, and file tasks and notes. 76 tools in all.

2 ways to connect

  • Sign in. You add the address, select Connect, sign in to Liteo, pick a workspace, and approve. This is how ChatGPT, claude.ai, and Claude Desktop connect. Claude Code, Codex, and Cursor can use it too.
  • API key. For Claude Code, Codex, Cursor, and your own scripts. You keep the key in an environment variable, and the client sends it with every call.

Either way, your AI works as a member of your workspace, with that member's permissions.

Connect ChatGPT

  1. In ChatGPT, turn on developer mode and add a new connector.
  2. Enter https://app.liteo.io/api/mcp as the URL and choose OAuth as the authentication.
  3. ChatGPT sends you to Liteo. Sign in, pick the workspace, and approve.

Connect claude.ai or Claude Desktop

  1. Open Settings, then Connectors, and choose Add custom connector.
  2. Enter https://app.liteo.io/api/mcp. Leave the OAuth client fields empty. Claude registers itself.
  3. Select Connect, sign in to Liteo, pick the workspace, and approve.

Liteo's approval screen shows which app is asking and where it sends you back. Check both before you approve.

Connect Claude Code, Codex, or Cursor

These clients can sign in the same way. Use an API key instead if you'd rather keep the connection in your own config.

Claude Code, from the terminal:

claude mcp add --transport http liteo https://app.liteo.io/api/mcp

Then run /mcp in Claude Code and authenticate.

Codex reads ~/.codex/config.toml. The CLI, the IDE extension, and the desktop app share it:

[mcp_servers.liteo]
url = "https://app.liteo.io/api/mcp"

Then run codex mcp login liteo.

Cursor reads ~/.cursor/mcp.json, or .cursor/mcp.json in a project:

{
  "mcpServers": {
    "liteo": { "url": "https://app.liteo.io/api/mcp" }
  }
}

With no key in the file, Cursor connects by sign-in.

Use an API key instead

Decide which kind of key the AI should use. Both are created in the same place.

  • Personal key. Acts as you and stops working when your access ends. Right for your own Claude Code, Codex, or Cursor.
  • Workspace key. Only admins can create one. New records land with the member you choose. It can't manage automations or forms.

Create the key in Liteo:

  1. Open Settings, then Integrations.
  2. Open For developers.
  3. Select Create key.
  4. Name it after the client, such as Claude Code or Cursor, and choose the type.
  5. For a workspace key, choose who new records should land with.
  6. Copy the key before you leave the dialog.

Save it as an environment variable named LITEO_MCP_API_KEY, then point your client at it.

Claude Code, in a project .mcp.json. The file holds no secret, so it's safe to commit:

{
  "mcpServers": {
    "liteo": {
      "type": "http",
      "url": "https://app.liteo.io/api/mcp",
      "headers": { "Authorization": "Bearer ${LITEO_MCP_API_KEY}" }
    }
  }
}

Codex, in ~/.codex/config.toml:

[mcp_servers.liteo]
url = "https://app.liteo.io/api/mcp"
bearer_token_env_var = "LITEO_MCP_API_KEY"

The Codex desktop app only sees variables from your login shell profile.

Cursor, in mcp.json:

{
  "mcpServers": {
    "liteo": {
      "url": "https://app.liteo.io/api/mcp",
      "headers": { "Authorization": "Bearer ${env:LITEO_MCP_API_KEY}" }
    }
  }
}

Check the connection

Ask your AI "Which Liteo workspace are you connected to?" It runs get_workspace and answers with the workspace name and the member it acts as.

If your client limits how many tools it shows, load a subset: https://app.liteo.io/api/mcp?toolsets=crm,operations. The toolsets are crm, operations, catalog, commerce, automations, and forms.

What the AI can and can't do

  • Any connection: search and read your CRM data, the catalog, sites, locations, installed products, service history, quotes, and invoices. Create and update contacts, businesses, deals, deal lines, sites, locations, installed products, and catalog items. Create notes and tasks, complete tasks, and add or remove contact tags. Record service visits and create draft quotes and invoices.
  • Admins only: read an automation in full, and create, change, turn on, or delete automations. Create, change, switch off, or delete forms. Liteo checks on every call that the person behind the connection is still an admin.
  • Needs confirmation on the call: archiving a site, removing an installed product, deleting an automation or a form. The call fails without it. Your AI app sees these as removals, so it can ask you before running them.
  • Never: sending, accepting, or issuing a quote or an invoice. Recording a payment, voiding, or credit notes. Setting up webhooks, putting a form live, or deleting a contact, business, or deal. Those stay in Liteo.

Reads work while your workspace has no active plan. Writes don't. The same request limits apply as for the REST API. Read the API documentation for toolsets, pagination, and limits.

Disconnect a client

Open Settings, then Integrations, and choose For developers.

  • For a client that signed in, find it under Connected apps and select Disconnect.
  • For a client that uses an API key, find the key and select Revoke key.

The client loses access right away. Nothing it already created is removed.

Still stuck?

Tell us what you expected and what happened. We’ll help you find the cause.

Email hello@liteo.io