Step 2: Connect Gmail and Calendar Services

Ensure all Gmail and Calendar nodes used by the agents and actions are authenticated.

1. Credential Required: Connect your gmailOAuth2 credentials to Clear Inbox Labels, Apply Gmail Label, Move to Trash, and Generate Reply Draft.
2. For tool access, Fetch Email Thread, Review Sent History, and Fetch Thread Context are connected as tools for AI agents; ensure Gmail OAuth2 credentials are available to the parent nodes (Email Classification Agent and Reply Drafting Agent).
3. Credential Required: Connect your googleCalendarOAuth2Api credentials for the tool node Calendar Availability Check (used by Reply Drafting Agent).

Step 3: Set Up Email Classification Agent

Configure the AI classification stage that determines labeling, deletion, and whether a response is needed.

1. Add Gemini Chat Model and connect it as the language model for Email Classification Agent and Reply Drafting Agent. Credential Required: Connect your googlePalmApi credentials to Gemini Chat Model.
2. In Email Classification Agent, keep Prompt Type set to define and ensure the prompt includes your label rules and placeholders (e.g., [YOUR_EMAIL] and label IDs).
3. Attach Structured JSON Parser as the output parser with the schema set to { "label": "Label name", "labelID": "Label ID", "id": "Email message ID", "requiresResponse": true }.
4. Confirm the tool connections: Fetch Email Thread and Review Sent History are linked as tools to Email Classification Agent.

Step 4: Configure Reply Drafting Agent

Set up the drafting logic that creates a draft reply when a response is required.

1. In Reply Drafting Agent, keep Prompt Type set to define and verify the prompt rules for tone, language detection, and scheduling.
2. Attach Draft JSON Parser as the output parser with schema { "draft": "string", "notes": "string" }.
3. Ensure tool connections are present: Fetch Thread Context and Calendar Availability Check are linked as tools to Reply Drafting Agent.
4. In Calendar Availability Check, confirm Calendar is set to [YOUR_EMAIL] and Timezone is set to [YOUR_TIMEZONE].

Step 5: Configure Output and Routing Logic

Define how emails are labeled, trashed, or drafted based on AI classification outputs.

1. In Clear Inbox Labels, set Operation to removeLabels, Message ID to ={{ $json.id }}, and include label IDs like INBOX, IMPORTANT, and category labels.
2. In Deletion Decision, set the condition to compare ={{ $json.output.label }} equals DELETE. The true branch goes to Move to Trash, and the false branch goes to Apply Gmail Label.
3. In Move to Trash, set Operation to addLabels, Label IDs to TRASH, and Message ID to ={{ $('Incoming Gmail Watch').item.json.id }}.
4. In Apply Gmail Label, set Operation to addLabels, Label IDs to ={{ $json.output.labelID }}, and Message ID to ={{ $('Incoming Gmail Watch').item.json.id }}.
5. In Response Requirement Check, set the condition to true when ={{ $('Email Classification Agent').item.json.output.requiresResponse }} is true.
6. In Generate Reply Draft, set Resource to draft, Message to ={{ $json.output.draft }}, Subject to =Re: {{ $('Incoming Gmail Watch').item.json.Subject }}, and Send To to ={{ $('Incoming Gmail Watch').item.json.From }}.

Step 6: Test and Activate Your Workflow

Validate that classification, labeling, and drafting behave as expected before turning the workflow on.

1. Click Execute Workflow and send a test email to the connected Gmail account.
2. Confirm that Clear Inbox Labels removes existing labels and that Email Classification Agent outputs valid JSON via Structured JSON Parser.
3. Verify that emails labeled DELETE move to trash, while others get labels from Apply Gmail Label.
4. For emails requiring a reply, check that Generate Reply Draft creates a draft with subject prefix Re: and body from {{$json.output.draft}}.
5. Once validated, switch the workflow to Active to run continuously.