Step 2: Connect Gmail and Telegram Services

Prepare the external services used for approval and sending replies.

1. Open Telegram Approval Prompt and set Chat ID to your Telegram user or group ID: =[YOUR_ID].
2. Confirm Operation is sendAndWait and Approval Type is double.
3. Credential Required: Connect your telegramApi credentials on Telegram Approval Prompt.
4. Open Dispatch Gmail Reply and keep the Gmail fields mapped to the AI output values.
5. Credential Required: Connect your gmailOAuth2 credentials on Dispatch Gmail Reply.

Step 3: Set Up the AI Decision Engine

Configure the AI to decide whether an incoming email needs a reply.

1. Open Validate Inbox Label and set the condition to check labelIds contains INBOX using {{ $json.labelIds }}.
2. Open AI Decision Engine and confirm the Text field includes the email context: =todays date:- {{ $now }} mail date:- {{ $json.headers.date }} from :- {{ $json.headers.from }} subject :- {{ $json.headers.subject.toJsonString() }} content :- {{ $json.text }}.
3. Ensure AI Decision Engine has Output Parser enabled.
4. Open Parse Decision JSON and keep the JSON Schema Example as { "response": "no" }.
5. Credential Required: Connect your openAiApi credentials on OpenAI Chat Engine.
6. Note: Parse Decision JSON is an AI sub-node connected to AI Decision Engine. Add credentials to OpenAI Chat Engine, not the parser.

Step 4: Set Up the AI Draft Reply

When a reply is required, draft the response using AI and structure the output.

1. Open Verify Reply Needed and confirm it checks {{ $json.output.response }} equals yes.
2. Open AI Draft Reply and confirm the Text field is set to: =todays date:- {{ $now }} mail date:- {{ $('Incoming Email Trigger').item.json.headers.date }} from :- {{ $('Incoming Email Trigger').item.json.headers.from }} subject :- {{ $('Incoming Email Trigger').item.json.headers.subject }} content :- {{ $('Incoming Email Trigger').item.json.text }}.
3. Ensure AI Draft Reply has Output Parser enabled.
4. Open Structured Parser A and keep the JSON Schema Example as { "email": "email address", "subject": "subject text", "body": " body text ." }.
5. Credential Required: Connect your openAiApi credentials on OpenAI Chat Engine 2.
6. Note: Structured Parser A is an AI sub-node connected to AI Draft Reply. Add credentials to OpenAI Chat Engine 2, not the parser.

Step 5: Configure Approval Routing and Reply Dispatch

Send the AI draft to Telegram, wait for approval, and conditionally send the Gmail reply.

1. Open Telegram Approval Prompt and review the Message template to include incoming email details and AI-generated reply: =📥 *New Email Received* 👤 *From:* {{ $('Incoming Email Trigger').item.json.headers.from }} 📝 *Subject:* {{ $('Incoming Email Trigger').item.json.headers.subject }} 💬 *Content:* {{ $('Incoming Email Trigger').item.json.text }} 🤖 *AI-Generated Response* 📤 *To:* {{ $('Incoming Email Trigger').item.json.headers.from }} 📝 *Subject:* {{ $json.output.subject }} 💬 *Content:* {{ $json.output.body }} .
2. Confirm Telegram Approval Prompt uses Limit Wait Time of 5 minutes.
3. Open Verify Telegram Approval and confirm it checks {{ $json.data.approved }} equals true.
4. Open Dispatch Gmail Reply and keep the fields mapped: Send To {{ $('AI Draft Reply').item.json.output.email }}, Subject {{ $('AI Draft Reply').item.json.output.subject }}, Message {{ $('AI Draft Reply').item.json.output.body }}, and Email Type text.
5. Review the no-op endpoints: End: No Reply for non-reply emails and End: Telegram Rejected when approval is denied.

Step 6: Test and Activate Your Workflow

Run a manual test to verify the full approval loop and email dispatch.

1. Click Execute Workflow and send a test email to the monitored Gmail inbox.
2. Confirm the email passes Validate Inbox Label and is analyzed by AI Decision Engine.
3. When a reply is needed, verify Telegram Approval Prompt sends the draft to Telegram and waits for approval.
4. Approve the message in Telegram and confirm Dispatch Gmail Reply sends the response.
5. If successful, toggle the workflow to Active to enable continuous monitoring.