How to build Any CRM with AI: Claude, ChatGPT or Codex on Make or n8n

Not sure how to build flows in Make or n8n? Connect Claude, ChatGPT or Codex to your automation tool and ask the AI to build the Any CRM webhook for you. This guide shows how to connect them, gives you a ready-made prompt and explains what to review before plugging everything into LinkedScope. The full format LinkedScope expects is in How to connect any CRM through n8n, Make or a webhook.

• Time: 15 to 30 minutes
• You’ll need: Claude, ChatGPT or Codex and a Make or n8n account
• Result: A webhook ready to paste into the Any CRM card
• AI connection: MCP, the standard that lets AI use external tools

How it works

1. You connect the AI to Make or n8n through each tool’s official MCP server.
2. You add the CRM connection and the webhook secret in the automation tool yourself.
3. You paste the prompt from this guide. The AI builds the flow, tests it and shows a sample response.
4. You review it, activate the flow and connect its URL in LinkedScope’s Any CRM card.

Before you start

• In LinkedScope: owner or admin access, a plan that includes CRM integration and no other CRM connected.
• In Make: a paid plan. The free plan only lets the AI run scenarios, not create or edit them.
• In n8n: version 2.13 or later and instance owner or admin access to turn on MCP.
• In the AI: a plan that supports custom connectors or apps (paid plans in ChatGPT), or Codex installed.
• An API credential for your CRM that can read deals, companies and users.

Don’t paste keys into the chat
Add the CRM key and the webhook secret directly on the connections and credentials screens in Make or n8n. Give the AI only the connection name, so no secret ends up in the conversation history.

Step 1: connect the AI to Make

MCP Toolboxes won’t do this
Make’s MCP Toolboxes screen only exposes finished scenarios for the AI to run. For the AI to build the scenario, connect the Make MCP server with management scopes, as below.

Claude

1. Open connectors
   In Claude, go to Settings > Connectors and click Browse connectors.
2. Add Make
   Search for Make, click + and sign in to your Make account.
3. Pick organization and scopes
   Select the organization and, besides Run your scenarios, check the management scopes, such as View and modify your scenarios. Without them, the AI can’t build the scenario.

ChatGPT

1. Use the Make app
   The simplest option is to install the official Make app from ChatGPT’s app directory and authorize it with the same management scopes.
2. Or add a custom connector
   In ChatGPT’s settings, turn on developer mode and create a custom app/connector with the URL https://mcp.make.com and OAuth authentication.

Codex

In a terminal, add the server and sign in through the browser. The setup also applies to the Codex editor extension and the ChatGPT desktop app:

codex mcp add make --url https://mcp.make.com
codex mcp login make

Step 1 (alternative): connect the AI to n8n

1. Turn on instance MCP
   In n8n, go to Settings > Instance-level MCP and click Enable MCP access.
2. Copy the connection details
   Click Connect a client, choose the OAuth (recommended) tab and select your client. The server URL ends in /mcp-server/http.
3. Connect Claude or ChatGPT
   In Claude, use Settings > Connectors > Add custom connector and paste the URL. In ChatGPT, create a custom connector with the same URL, with developer mode on. Both sign in through OAuth.
4. Or connect Codex
   Run the install command n8n shows, or the commands below with your instance’s address.

codex mcp add n8n --url https://your-n8n.example.com/mcp-server/http
codex mcp login n8n

For the AI to test a flow over MCP after building it, the flow must be published with Available in MCP turned on (the flow’s … menu > Settings).

Step 2: add the credentials

• CRM connection: in Make, under Connections; in n8n, under Credentials. Note the name you give it.
• Webhook secret: generate a long random value. In n8n, create a Header Auth credential with the header name auth and that value. In Make, the key is generated in the webhook’s own API key authentication, after the AI builds the scenario.

Step 3: paste the prompt

Replace the bracketed parts and send it to the AI:

Build a flow in my [Make | n8n] that works as LinkedScope's "Any CRM" integration for my CRM, [CRM NAME].

Requirements:
1. Trigger: a webhook that receives a POST with no body.
   - Make: a Custom webhook module with API key authentication (x-make-apikey header) and a Webhook response module at the end.
   - n8n: a Webhook node, POST method, Header Auth authentication with the "[WEBHOOK CREDENTIAL NAME]" credential (auth header), responding through the Respond to Webhook node.
2. Using the existing "[CRM CONNECTION NAME]" connection/credential, fetch every deal created or closed in [CRM NAME] in the last 24 months. Paginate to the end; don't stop at 100 records.
3. For each deal, get the company (name, website domain and, if available, LinkedIn page URL) and the owner (name and email). Fetch companies and users in bulk, not one call per deal.
4. Build a JSON array where each item has: dealId, dealName, dealLink, pipeline, dealStage, dealValue (number), createdAt and closedAt (ISO 8601), isWon (true if won, false if lost, null if open), ownerName, ownerEmail, companyName, companyDomain and linkedinCompanyUrl. Empty fields are null.
5. Respond only at the end of the flow, with HTTP 200, Content-Type application/json and the array.
6. Don't write anything to the CRM and don't type keys or tokens into modules: use only the existing connections and credentials.

Before building, tell me which [CRM NAME] endpoints you'll use and how you'll map stage, amount and status. After building, run a test and show me 3 items from the response.

If your CRM has anything specific, such as several pipelines or the amount in another field, say so in the prompt. To analyze a single business unit, ask the flow to filter deals before responding.

Step 4: review and test

Before connecting it to LinkedScope, check the flow:

• The trigger accepts POST and rejects calls without the right secret (401 or 403).
• The fetch paginates to the end and covers the last 24 months.
• The response comes from the last step, with the deal array. Make needs a Webhook response module; n8n, a Respond to Webhook node.
• The flow is active: in Make, the scenario is on and set to run immediately; in n8n, the flow is published and you use the Production URL (not the test one).

To test it yourself, run this in a terminal with the real URL and secret. The response should be a JSON array:

curl -X POST "https://WEBHOOK-URL" -H "auth: YOUR-SECRET"

Step 5: connect in LinkedScope

In Settings > Integrations, open the Any CRM card, paste the webhook URL into Webhook trigger URL and only the secret into Authentication token. LinkedScope sends the auth header and, if it is rejected, retries with x-make-apikey, so the same field works for n8n and Make. Save and click Test. Field and sync details are in How to connect any CRM through n8n, Make or a webhook.

Common errors

Frequently asked questions

Which AI should I use?

Any of the three works. Claude and ChatGPT are easier if you don’t code. Codex suits people who already work in a terminal or editor.

Does the flow need the AI to run?

No. Once built, the flow runs on its own in Make or n8n. You can even remove the AI connector.

Does the AI see my CRM data?

While building and testing, the AI sees what Make or n8n tools return, such as the deal sample. LinkedScope sends no data to the webhook: the call is a POST with no body.

Can I ask for changes later?

Yes. With the connector active, ask the AI to edit the existing flow, for example to add a pipeline or fix a field, and test again in the Any CRM card.

Need help?

Reach us at [email protected], through the Help menu inside LinkedScope or on our contact page.