Now in ChatGPT & ClaudeConnect

Connect Chatfuel MCP with Gmail MCP

When customers in WhatsApp or Instagram ask for details by email, a personal draft for each of them waits in your Gmail, ready to check and send. From the same assistant chat, the answers your team already wrote by email go into the AI Agent's FAQ, and a customer's email thread helps you reply in her chat. Google's Gmail server is in Developer Preview and only writes drafts, so every email leaves when you press Send.

What the integration can do

Tools the assistant calls on each side, named as the servers name them.

Chatfuel MCP
ToolWhat it does
list_chatsLists customer chats with their channel, status and lead stage.
get_chatReturns the latest messages of one customer chat with its channel, status and lead stage.
update_automation_settingChanges one setting of an automation: when the AI replies, handoff to your team, follow-ups, lead qualification, booking rules or comment replies.
review_chatsSummarizes customer chats for a period instead of returning every message.
get_knowledge_baseReturns what the AI Agent answers from: company info, instructions, business hours, FAQ, services and specialists.
add_faqAdds question and answer pairs to the FAQ the AI Agent answers from.
test_replySends a sample customer message to a sandbox chat and returns the AI Agent's reply.
send_chat_messageSends the text you approved to a customer in their chat and moves the chat to live chat.
return_chat_to_aiHands a chat in live chat back to the AI Agent after your yes, and the AI Agent answers the customer's last message.
list_bookingsLists appointments between two dates: time, service, specialist, customer name and status.
Gmail MCP
ToolWhat it does
search_threadsSearches your mailbox with Gmail search operators such as from:, to:, in:sent or newer_than: and returns threads with their subjects and short snippets.
get_threadReturns every message of one email thread with its sender, recipients, date and full text.
create_draftSaves a new email in your Drafts with recipients, subject and text, or as a reply to an existing message. It never sends anything.
list_draftsLists the drafts in your account, optionally filtered by a search such as to: and an address.
list_labelsLists the labels you created in Gmail with their IDs, which the other label tools and label: searches need.
label_threadAdds one or more labels to a whole thread, including messages that arrive in it later.

What to ask the assistant

Asked for details by email in the chat → a Gmail draft per customer

WhenPeople ask the AI Agent about the package of 3 Hydration facials ($159) and want the details by email, to read later or show a partner. Those emails rarely get written: they slip between chats.

  1. Chatfuel MCP turns on contact collection with update_automation_setting: when a customer wants the details by email, the AI Agent asks for the address and saves it to an email attribute.
  2. At the end of the day, list_chats lists the recent chats and get_chat reads them. The assistant keeps those where a customer typed an email and asked about the package.
  3. Gmail MCP checks with search_threads, using to: and that address, whether someone on your team already wrote to them, and leaves those people out.
  4. create_draft writes one email per customer: the package price, what the three visits include, how to book, and a line about what they asked in the chat.

ResultYour Drafts folder holds a personal email for each customer, and you send each one from Gmail after a read. The Gmail server has no send tool, so nothing reaches anyone before you click Send. Google's reference says drafts can't carry attachments yet, so link to the price list in the text.

Ask

Find today's Chatfuel chats where a customer left an email and asked about the package of 3 Hydration facials. Skip anyone we've already emailed, and draft a personal Gmail email for each of the others.

Chatfuel chat: a customer asks for the Hydration facial package details by email and types her address for the AI Agent

Your team's email answers → the AI Agent's FAQ

WhenSome customers write to the clinic's Gmail instead of the chat, and Grace Kim answers them there: invoices for companies, facials for teenagers. In WhatsApp and Instagram, the AI Agent still doesn't know those answers.

  1. Chatfuel MCP runs review_chats over the last month; among other things, it lists the customer questions no FAQ entry covers.
  2. Gmail MCP looks for each topic in your sent mail with search_threads (in:sent newer_than:90d), and get_thread reads how Grace answered.
  3. add_faq turns the answers you approve into FAQ entries, and test_reply asks the AI Agent one of the questions in a sandbox chat.
  4. list_labels finds the In FAQ label you made in Gmail, and label_thread puts it on the threads you used, so the next search skips them with -label.

ResultThe AI Agent answers in the chat with the words Grace already chose by email, and your mailbox shows which threads made it into the FAQ. Labels need the gmail.modify scope, which Google's setup page doesn't add.

Ask

Find the questions our Chatfuel FAQ doesn't cover yet, look up how Grace answered them in our sent Gmail over the last 90 days and add those answers to the FAQ. Then label the threads you used In FAQ.

Chatfuel FAQ with two new answers taken from the team's emails: invoices for companies and facials for teenagers

"I emailed you" in the chat → an answer from the email thread

WhenA customer writes on Instagram that she sent photos of her skin to the clinic's email last week and asks what Grace thinks. The AI Agent can't see your inbox.

  1. Chatfuel MCP adds a handoff rule with update_automation_setting: when a customer refers to an email she sent, the AI Agent says the team will check and passes the chat on.
  2. list_chats shows the chats waiting for a person, and get_chat reads her message and the address she typed.
  3. Gmail MCP finds her thread with search_threads, using from: and that address, and get_thread reads her email and Grace's reply: start with a Hydration facial, not a Peeling.
  4. send_chat_message sends the short answer you approve, and return_chat_to_ai hands the chat back after your yes, so the AI Agent can book her visit.

ResultShe gets Grace's advice in the chat she wrote from, and nobody copies text between apps. Instagram accepts your reply within 7 days of her last message, WhatsApp within 24 hours. If she typed no address, the assistant asks you which thread is hers.

Ask

Find the Chatfuel chats waiting for us where a customer mentions an email she sent. Read her thread in Gmail with Grace's reply and draft a short answer for the chat. Hand the chat back to the AI Agent once I've sent it.

Chatfuel chat: the AI Agent hands the chat to the team, which replies with Grace Kim's advice from the email thread

Yesterday's visits → aftercare emails that match the chat

WhenAfter a Peeling or a Hydration facial, clients ask the same aftercare questions in the chat, and Grace wants each of them to get written care tips by email the next morning.

  1. Chatfuel MCP lists yesterday's appointments with list_bookings and keeps the Peelings and Hydration facials marked as attended: time, service, specialist and client name.
  2. get_knowledge_base reads the aftercare answers in the FAQ. Any that are missing, add_faq adds after your yes, so the chat and the email say the same thing.
  3. list_chats finds each client's chat by name, and get_chat looks for the email she typed when she booked.
  4. Gmail MCP checks list_drafts for an aftercare draft to that address, then create_draft writes one per client with the tips for her treatment.

ResultClients who left an email get care tips in your words once you send the drafts, and the AI Agent gives the same answers when they ask in the chat. Chatfuel bookings carry no email, so clients who never typed one are listed for you instead.

Ask

Check yesterday's attended Peelings and Hydration facials in Chatfuel, make sure the FAQ has aftercare answers for both, and draft an aftercare email in Gmail for each client who typed her email in the chat.

Chatfuel FAQ with two new aftercare answers: redness after a Peeling and makeup after a Hydration facial

How to connect

  1. Get into Google's preview

    Gmail's MCP server answers only accounts in the Google Workspace Developer Preview Program. Apply with your Workspace account and a Cloud project; Google registers both and confirms by email, usually in a couple of days.

  2. Set up the Cloud project

    Enable the Gmail API and the Gmail MCP API, add the gmail.readonly and gmail.compose scopes on the consent screen (plus gmail.modify for labels) and create a Web application client marked Use this client for an AI-powered agent.

  3. Add both servers and ask

    Chatfuel comes from the Claude or ChatGPT directory, or from https://mcp.chatfuel.com/panel in Cursor. Gmail goes in by https://gmailmcp.googleapis.com/mcp/v1 with your client ID and secret. Chatfuel asks for a yes before each bot change and each customer message.

  1. Connect Chatfuel from Claude's connector directory and sign in.
  2. Claude's directory also has a Gmail connector made by Google, but Claude's help page says it only searches and reads mail. For drafts and labels, add Gmail as a custom connector.
  3. Give your Google OAuth client the redirect URI https://claude.ai/api/mcp/auth_callback. Then open Settings › Connectors (Admin settings › Connectors on Team and Enterprise), click Add custom connector, name it Gmail and paste https://gmailmcp.googleapis.com/mcp/v1.
  4. Enter the client ID and secret under Advanced settings, click Add and sign in with the Google account that's in the preview. Google lists Claude Pro, Max, Team and Enterprise for this.
Gmail MCP
https://gmailmcp.googleapis.com/mcp/v1
Chatfuel MCP in Claude

What you need

  • A Chatfuel bot whose AI Agent answers customers on WhatsApp, Instagram, Facebook, TikTok or website chat. The aftercare use case also needs booking set up, so the AI Agent books visits.
  • A Google Workspace account and Cloud project accepted into the Google Workspace Developer Preview Program, which Gmail MCP requires.
  • In that project: the Gmail API and the Gmail MCP API enabled, an OAuth consent screen, and a Web application OAuth client with Use this client for an AI-powered agent selected. Google's Workspace MCP servers have refused unmarked clients since November 11, 2026.
  • The gmail.readonly and gmail.compose scopes for reading mail and writing drafts, and gmail.modify for the label steps.
  • Claude on Pro, Max, Team or Enterprise, ChatGPT with Developer mode, or Cursor, signed in to the Gmail account your team answers customers from.

Limits

  • Gmail MCP is a Developer Preview. Google's program terms provide it as is, allow no users outside your company before GA, and let Google use data sent through it to improve the API.
  • Google documents no tool that sends email. The assistant writes drafts, and someone on your team sends each one from Gmail.
  • Per Google's create_draft reference, drafts can't include attachments yet. Link to a price list or care sheet in the email text instead.
  • Chatfuel gives names, not emails, for chats and bookings, and list_chats can't search by email. An address exists only if the customer typed it in the chat, and matching a chat to a thread relies on that address.
  • Claude's Gmail connector from its directory is read-only, per Claude's help page. Drafts and labels need the custom connector with your own OAuth client.
  • Emails can carry hidden instructions. Google tells you to screen prompts and responses for injection, with Model Armor or a tool of your own, and to review every action the assistant takes. Without offline access, Google also asks you to sign in again about every hour.

FAQ

  • Yes. Google runs it at https://gmailmcp.googleapis.com/mcp/v1 over Streamable HTTP as one of the Google Workspace MCP servers, labeled Developer Preview. It isn't in the MCP Registry and isn't generally available yet.

Connect Chatfuel MCP

Start a free trial, add Chatfuel to your assistant and ask about Gmail and your chats in one place.