How to install Laiki's MCP in Claude and ChatGPT

This tutorial installs the Laiki MCP connector in Claude or ChatGPT. With it, you query your account's leads, events, segments, campaigns and dashboards right in the conversation, without exporting spreadsheets and without leaving the AI agent you already use every day.

Before you start

You will need two things: a Claude or ChatGPT plan that allows custom connectors, and the login credentials for your Laiki account. The process takes a few minutes and is done only once per device.

MCP (Model Context Protocol) is the standard that lets an AI agent talk to external systems. What the Laiki connector can do depends on the permissions you check inside the platform: each resource has separate read and edit controls, and nothing is enabled without your action.

It is worth understanding why this differs from simply pasting a spreadsheet into the chat. A pasted spreadsheet is a snapshot of one moment, and someone has to export and update that file every time the question changes. The connector queries the source directly, so the answer reflects the real data in the account at the moment you ask, without that preparation work.

Enable MCP access in your Laiki account

The installation steps in Claude or ChatGPT connect the agent to the MCP URL, but the access itself stays off until you enable it inside the Laiki account. That is where you decide which data the agent can query and which it can edit.

Go to Credentials → AI Agents → MCP and turn on the MCP enabled switch. That same screen controls the allowed operations: each resource (Leads, Events, Insights, Segments, Custom fields, Context documents, Dashboards, Workflows, Credentials metadata) has its own Read permissions and, where applicable, Edit permissions. Enable edit access only on the resources where it is really necessary.

MCP is in beta: the available tools may still change.

Install in Claude

The steps below work in both the Claude desktop app and claude.ai in the browser; the Settings screen is the same in both environments.

1. Open Claude's Settings. Click your account name in the lower left corner of the screen and select Settings from the menu that opens.
2. Go to the Connectors tab. In the left column of Settings, inside the Customize block, click Connectors . This is where all of Claude's integrations live, both the prebuilt and the custom ones.
3. Click Add → Add custom connector. At the top right of the Connectors screen, open the Add button and choose the second option, Add custom connector . The Browse connectors option lists only ready-made integrations, and Laiki is not on that list, which is why the custom path is the way to go.
4. Fill in the name and the server URL. In the first field, type the name you want to see in the connector list (we recommend Laiki). In the second field, paste exactly this URL: https://mcp.laiki.co/mcp . Leave the Advanced settings fields (OAuth Client ID and OAuth Client Secret) blank, they are optional and are not used in the standard Laiki connection. Click Add to finish.
5. Authenticate and adjust the permissions. After you add it, Claude opens the Laiki login screen in the browser. Sign in with your user and authorize access. Once that is done, the connector appears in the list already linked. On the connector's screen you control what Claude can do: read-only tools can be set to Always allow , so queries run without interruption on every call. Needs approval makes Claude ask for your confirmation each time it uses the tool, useful while you are getting to know the integration, but in long analyses it generates many clicks. Blocked turns the tool off completely.

Install in ChatGPT

1. Open Settings and go to Apps & Connectors. Click your account icon in the lower left corner of the screen and select Settings .
2. Turn on Developer Mode. Inside Apps & Connectors, scroll to the bottom and click Advanced Settings . Turn on the Developer Mode option , a required step to add custom connectors in ChatGPT. If you use the ChatGPT desktop app, this setting has to be done first in the web version.
3. Go back to Connectors and click Create. The button sits next to Browse Connectors , on the same screen.
4. Fill in the name, description and server URL. Use Laiki as the name, a short description such as “Access to Laiki leads, events, segments and dashboards”, and paste the same connector URL: https://mcp.laiki.co/mcp .
5. Authenticate and confirm. ChatGPT opens the Laiki login screen for OAuth authentication. Sign in with your user and authorize access. Check the option to trust the app when prompted, and click Create .
6. Use it in the conversation. In the message field, click the + icon and select the Laiki connector from the list. From then on, any question in the conversation can query your account's data.

Already had the connector installed with a different address? Remove the old connector and add it again with https://mcp.laiki.co/mcp. The login and permissions are redone in the same flow above.

After installing

Confirm that the connection is up. Ask the agent: “list my organizations in Laiki”. If the list of accounts comes back, everything is fine.

You do not need to provide the account ID. The agent resolves it on its own on the first query, when it lists your organizations. The only situation where it is worth saying up front which account you want to work with is if your access has several organizations registered, so it does not pick the wrong one.

The account context travels along. Funnel, event names, custom fields and UTM convention are stored as context documents inside the account itself, and the agent reads this material automatically when it opens the organization. You do not need to have any document at hand or attach a file to each conversation. If your account does not have this context registered yet, talk to the Laiki team.

Start with a simple question. Examples: “how many leads came in this month?”, “how is the funnel doing over the last 30 days?”, “what is the CPL by channel?”. More specific questions, such as crossing segments with media attribution, work better after the connection has been validated with a simple example.

Repeat this process on each device. The connector is tied to the environment where it was installed. If you use Claude or ChatGPT on your office computer and also in the browser at home, repeat the steps on both; there is no single installation that covers all devices at once.

If something does not work

An error when adding the connector is usually a URL with an extra space or a wrongly pasted character. Type the URL manually instead of copying it from a formatted document, which sometimes inserts invisible characters along with the text.

Connector linked but not responding. In Claude, use Disconnect on the connector's screen and log in again. In ChatGPT, remove the connector in Apps & Connectors and add it again. In most cases, this restarts the authentication session and fixes it.

MCP returns no data even with the connector linked. Check that MCP is enabled in Credentials → AI Agents → MCP, and that the required operations (read or edit) are checked for the resource you are trying to query.

If it persists, talk to the Laiki team at laiki.co.

What comes after the installation

With the connector active, the natural next step is knowing what to ask. The Laiki Prompt Guide offers 70 ready-to-copy prompts, organized by area and by data source, designed exactly for those who just installed the connector and are still discovering what to ask. To better understand how MCP works behind those questions, the articles on what MCP is and how Laiki's MCP works detail the mechanism and the real resources behind the connection.