Connect Kustomer to ChatGPT: Manage Profiles and Conversation History

You want to connect Kustomer to ChatGPT so your AI agents can read customer profiles, summarize conversation histories, and draft omnichannel replies based on historical context. If your team uses Claude instead, check out our guide on connecting Kustomer to Claude, or explore our broader architectural overview on connecting Kustomer to AI Agents.

Giving a Large Language Model (LLM) read and write access to your Kustomer instance is an engineering challenge. Kustomer is not a flat ticketing system. It is a complex, timeline-based CRM heavily reliant on custom objects and omnichannel threading. You either spend weeks building, hosting, and maintaining a custom Model Context Protocol (MCP) server, or you use a managed infrastructure layer that handles the boilerplate for you.

This guide breaks down exactly how to use Truto to generate a secure, managed MCP server for Kustomer, connect it natively to ChatGPT, and execute complex support and data extraction workflows using natural language.

The Engineering Reality of the Kustomer API

A custom MCP server is a self-hosted integration layer that translates an LLM's tool calls into REST API requests. While Anthropic's open MCP standard provides a predictable way for models to discover tools, the reality of implementing it against vendor APIs is painful. If you decide to build a custom MCP server for Kustomer, you are responsible for the entire API lifecycle.

Here are the specific integration challenges that break standard CRUD assumptions when working with Kustomer:

The KObject (Custom Object) Klass System

Most ticketing platforms have a fixed schema. Kustomer relies on "Klasses" - dynamic custom objects that define everything from internal orders to external shipping data. If an LLM needs to look up a customer's recent return, it cannot hit a standard /returns endpoint. It has to query the list_all_kustomer_custom_objects endpoint and pass the specific klass name. Your MCP server must be able to dynamically read the schema of these custom Klasses so the LLM understands what fields are available to filter or update. Hardcoding JSON schemas for Kustomer custom objects guarantees your integration will break the moment a support admin adds a new field.

Timeline Unmasking and PII

Kustomer takes data privacy seriously. By default, sensitive attributes in customer profiles, conversations, and custom objects may be masked. If your AI agent needs to read a masked phone number or a sensitive custom field to perform identity verification, it cannot just call a GET endpoint. The integration layer must explicitly hit the create_a_kustomer_customer_unmasking_window endpoint to open a temporary window where that specific customer's data is returned unmasked. Failing to account for this means your LLM operates on redacted data and hallucinates the missing context.

Omnichannel Draft Fragmentation

In Kustomer, drafting a reply isn't a single universal action. The API requires you to specify the exact channel context. There are completely different endpoints and payload structures for creating an Email draft (create_a_kustomer_email_draft), an SMS draft (create_a_kustomer_customer_sms_draft), a Twitter DM draft, or a WhatsApp interactive draft. Your MCP server must expose discrete, well-documented tools for each channel, otherwise, the LLM will attempt to push an HTML-formatted email payload into an SMS endpoint, resulting in a 400 Bad Request.

Rate Limits and 429 Errors

Kustomer enforces strict rate limits, including both global limits and object-level limits (e.g., you can only create 120 conversations per minute per customer). When you hit these limits, Kustomer rejects the request. Truto does not retry, throttle, or apply backoff on rate limit errors. When the upstream Kustomer API returns an HTTP 429, Truto passes that error directly to the caller, normalizing the upstream rate limit information into standardized IETF headers (ratelimit-limit, ratelimit-remaining, ratelimit-reset). The caller (your LLM agent framework) is entirely responsible for reading these headers, pausing execution, and applying exponential backoff.

Generating a Kustomer MCP Server

Instead of forcing your engineering team to build a custom server that manually maps Kustomer's Klass schemas to JSON-RPC endpoints, you can use Truto to dynamically generate a managed MCP server.

Truto reads the Kustomer API documentation and your specific environment's custom object definitions, automatically translating them into MCP-compatible tools. Each MCP server is backed by a cryptographic token that routes requests to the correct integrated account.

You can create this MCP server via the Truto UI or programmatically via the REST API.

Method 1: Creating the Server via the Truto UI

For internal tooling and fast prototyping, the UI is the quickest path:

Log into your Truto dashboard and navigate to the Integrated Accounts page.
Select your connected Kustomer instance.
Click the MCP Servers tab.
Click Create MCP Server.
Select your desired configuration (e.g., filtering for specific tags or limiting to read-only methods).
Copy the generated MCP Server URL (e.g., https://api.truto.one/mcp/a1b2c3d4e5f6...).

Method 2: Creating the Server via the REST API

For production use cases - like generating isolated MCP servers for individual users - you should use the API. This endpoint generates a secure token stored in Cloudflare KV and returns a ready-to-use URL.

curl -X POST https://api.truto.one/integrated-account/<YOUR_KUSTOMER_ACCOUNT_ID>/mcp \
  -H "Authorization: Bearer <YOUR_TRUTO_API_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Support Ops AI Agent",
    "config": {
      "methods": ["read", "write", "custom"]
    },
    "expires_at": "2026-12-31T23:59:59Z"
  }'

The API returns a JSON payload containing the server URL:

{
  "id": "mcp-789-xyz",
  "name": "Support Ops AI Agent",
  "config": { "methods": ["read", "write", "custom"] },
  "expires_at": "2026-12-31T23:59:59.000Z",
  "url": "https://api.truto.one/mcp/a1b2c3d4e5f67890"
}

This single URL is all your LLM framework needs to authenticate, discover tools, and execute API calls against Kustomer.

Connecting the MCP Server to ChatGPT

Once you have your Truto MCP URL, you need to register it with ChatGPT. You can do this natively via the ChatGPT UI for Enterprise/Pro accounts, or by configuring your agent framework's JSON file.

Method A: Via the ChatGPT UI

If you are using ChatGPT Enterprise or Pro with developer features enabled, you can connect the server directly:

Open ChatGPT and navigate to Settings -> Apps -> Advanced settings.
Toggle Developer mode on (MCP support requires this flag).
Under the MCP servers / Custom connectors section, click Add a new server.
Name: Enter a descriptive label (e.g., "Kustomer Production Data").
Server URL: Paste the Truto MCP URL generated in the previous step.
Click Save.

ChatGPT will immediately ping the server's initialize endpoint, perform the handshake, and request the tools/list. Your Kustomer tools are now available to the model.

Method B: Via Manual Config File

If you are running a local instance of Claude Desktop or a custom agent framework that consumes MCP config files, you can map the Server-Sent Events (SSE) URL directly.

{
  "mcpServers": {
    "kustomer-truto": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sse",
        "--url",
        "https://api.truto.one/mcp/a1b2c3d4e5f67890"
      ]
    }
  }
}

Because Truto embeds the authentication and routing context directly into the cryptographic token in the URL, you do not need to inject Kustomer API keys or OAuth tokens into your local environment variables.

Security and Access Control

Handing an LLM a Kustomer API key is reckless. Models are prone to hallucinating parameters, which can lead to destructive updates or mass deletions. Truto's MCP architecture provides strict access control layers to sandbox the AI agent.

Method Filtering: When generating the server, you can pass config.methods: ["read"]. The LLM will only be able to see and execute get and list endpoints. Write operations (create, update, delete) are entirely stripped from the tool list.
Tag Filtering: You can restrict the server to specific Kustomer resources using config.tags. For example, setting tags to ["conversations"] prevents the LLM from accessing billing data, custom objects, or routing rules.
Require API Token Auth: By enabling require_api_token_auth: true, possession of the MCP URL is no longer enough. The client must also pass a valid Truto API token in the Authorization header, preventing unauthorized access if the URL leaks.
Automatic Expiration: Setting an expires_at timestamp binds the server to a Cloudflare KV TTL and Durable Object alarm. Once the timestamp passes, the server ceases to exist, making it perfect for temporary, task-specific agent sessions.

Hero Tools for Kustomer

Truto automatically generates dozens of tools based on the Kustomer schema. Instead of overwhelming the LLM with every possible REST endpoint, here are the highest-leverage tools for building Support and RevOps agents.

1. kustomer_customers_using_email

Look up a customer profile using their email address. This is almost always the first step in a support workflow, as you need the internal Customer ID to fetch related conversations, custom objects, or notes.

"Find the customer record for jane.doe@example.com and tell me what their internal ID is, along with their current VIP status if it exists in their attributes."

2. list_all_kustomer_conversations

Fetches a paginated list of conversations. When dealing with support requests, the LLM uses this tool to check if the user already has an open thread regarding their issue.

"List the 5 most recent conversations for customer ID 64a1b2c3d4e5f6. Are any of them currently marked as 'open' or 'snoozed'?"

3. create_a_kustomer_message

Adds a message to an existing conversation timeline. This requires the conversation ID, message type, channel, and the app context. The LLM uses this to append internal context or draft actual replies.

"Add a message to conversation ID 99x88y77z. Set the channel to 'email' and include this summary: 'Customer requested a refund due to late shipping. Escalating to billing team.'"

4. get_single_kustomer_sla_by_id

Retrieves SLA policy details. If an agent is deciding which ticket to prioritize, it can fetch the SLA rules to determine breach thresholds.

"Fetch the SLA details for SLA ID 12345. What is the designated response time threshold for VIP customers on this policy?"

5. kustomer_custom_objects_using_external_id

Retrieves a KObject (custom object) using a specific klass name and an external identifier. This is critical for connecting external inventory, shipping, or subscription data stored in Kustomer back to the user.

"Fetch the custom object with klass name 'order' and external ID 'ORD-98765'. What is the current shipping status of this record?"

6. create_a_kustomer_customer_unmasking_window

Allows the LLM to request temporary unmasking of sensitive attributes for a specific customer. This is vital when the agent needs to verify identity or process data that Kustomer natively redacts.

"Create an unmasking window for customer ID 64a1b2c3d4e5f6 so I can read their exact phone number. Once unmasked, output the E.164 formatted number."

For the complete inventory of available Kustomer tools - including bulk operations, team management, and draft endpoint schemas - view the Kustomer integration page.

Workflows in Action

Tools are just discrete operations. The real power of MCP is how ChatGPT chains these tools together to execute multi-step workflows without human intervention.

Workflow 1: The Support Triage Agent

Instead of a support manager manually reading an incoming email, looking up the user, checking their history, and drafting a response, ChatGPT handles the entire loop.

"A user just emailed from alex.smith@acmecorp.com complaining that their recent order hasn't arrived. Find their account, check their recent conversations to see if they've asked about this before, look up their custom 'order' object for the status, and then add a private note to their active conversation summarizing what you found."

Step-by-step Execution:

Tool Call: kustomer_customers_using_email - The LLM searches for alex.smith@acmecorp.com and extracts the Customer ID (cust_123).
Tool Call: list_all_kustomer_conversations - The LLM queries recent threads for cust_123 to establish context and finds an open conversation (conv_456).
Tool Call: kustomer_custom_objects_using_external_id - The LLM queries the klass named order using the customer's external ID to check the fulfillment status.
Tool Call: create_a_kustomer_conversation_note - The LLM calls the note endpoint, appending a private summary to conv_456 stating that the order is delayed in transit and no previous complaints exist.

Workflow 2: Automated Profile Sync and Unmasking

When syncing data from a backend database to Kustomer, you often need to merge profiles or update sensitive contact details.

"We need to update the phone number for the user with external ID 'USR-555'. First, look them up by external ID. If their current phone number is masked, create an unmasking window to verify it. If the unmasked number does not match '+14155552671', update their profile with the new number."