Trip plan email with flights, hotels and activities

Trip plan email with flights, hotels and activities

An n8n workflow that answers one web request with a travel plan by email. Another app, a form or a voice agent posts the
cities, the dates, the number of travelers and the kind of things to do. Claude turns the cities into airport codes,
searches the web for activities, and SerpApi finds flights on Google Flights and hotels on Google Hotels. Claude then
writes an HTML email with the three lists, Gmail sends it, and the request gets a one-line confirmation back.

What is inside

• workflow.json: the workflow as n8n exports it, fourteen nodes and no sticky notes. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• Send Plan sent every plan to one personal address. It now sends to YOUR_EMAIL.
• The email prompt signed off with a travel brand's name. It now says YOUR_TEAM_NAME.
• All three Claude steps named claude-sonnet-4-20250514. Anthropic's deprecations page, read on 2 November 2026, lists
  that model as retired on 15 June 2026, so requests to it fail. They now use claude-sonnet-5-5, the replacement that
  page names.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An Anthropic API key. Activities uses Claude's web search, which is on unless an admin turned it off in the Claude
  Console.
• A SerpApi account and its API key.
• A Gmail account to send from.
• Something that can send a POST request with a JSON body: another workflow, a form tool, a voice agent or curl.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add an Anthropic credential and select it in Anthropic Chat Model, Anthropic Chat Model1 and Activities.
3. Add a Query Auth credential for SerpApi: the parameter name is api_key and the value is your key. Select it in
   Flights and Resorts.
4. Add a Gmail OAuth2 credential and select it in Send Plan. Replace YOUR_EMAIL in Send to with the inbox for the
   plans. To mail each traveler instead, put {{ $('Set Fields').item.json.email }} there.
5. In the system message of Email Agent, replace YOUR_TEAM_NAME with the name the email should be signed with.
6. Save. Copy the test address from Webhook, send it the request below, then publish the workflow and use its
   production address.

The request

Webhook takes a POST at the path travel. Set Fields reads these fields from the JSON body. Spell the departure date
depature_date, as the file does, or that field arrives empty.

{
  "origin": "Chicago",
  "destination": "Lisbon",
  "depature_date": "2026-12-12",
  "return_date": "2026-12-19",
  "travelers": 2,
  "activities": "food tours and day trips",
  "email": "you@example.com"
}

The answer comes back as JSON once the email is out. Its description holds the sentence from Response, and the rest
of it looks like the settings of a voice agent's end-call tool, so a voice agent can read the result and hang up.

How a run goes

1. Set Fields takes the seven values out of the request body.
2. Airport Codes & Dates has Claude turn both cities into airport codes and check that the dates lie in the future,
   against the current time. Locations & Dates holds its answer to origin, destination, departure and return.
3. Activities asks Claude to research things to do of the requested kind at the destination, with up to five web
   searches.
4. Resorts searches Google Hotels through SerpApi for the destination, the dates and the number of adults, sorted by
   highest rating.
5. Flights searches Google Flights through SerpApi between the two airport codes on those dates.
6. Email Agent receives two flights, three hotels and three activities in a fixed text and writes the email as HTML
   in four parts: an introduction, flights, hotels and activities. Subject & Email holds its answer to subject and
   emailBody.
7. Send Plan sends it as an HTML email, Response writes the confirmation sentence and Respond to Webhook returns it.

Change the look

• Sections, tone and length: the system message of Email Agent. It asks for a horizontal line after each section,
  hotel pictures at a fifth of the width and no more than 1,000 words.
• How many results the email names: the text of Email Agent, which lists flights [0] and [1], and hotels and
  activities [0] to [2].
• Hotel order: sort_by in Resorts. SerpApi's Google Hotels documentation gives 3 for lowest price,
  8 for highest rating and 13 for most reviewed.
• Activity research: the prompt of Activities, and its search limit of 5.
• Model: claude-sonnet-5-5 in Anthropic Chat Model, Anthropic Chat Model1 and Activities.

Good to know

• Nightly rates: the first and second hotels show the third hotel's rate per night and total, because their lines in
  the Email Agent text read properties[2]. Change those two to properties[0] and properties[1].
• Activities: Email Agent reads them from the raw search results in content[1] of the Activities answer, which
  holds them only when Claude searches before it writes any text. Their description is the search result's
  encrypted_content, an encrypted string, so Claude writes the descriptions from the titles. The JSON list the
  Activities prompt asks for is never read.
• Fewer results fail: the text expects two flights, three hotels with two nearby places each and three activities. A
  search that returns fewer stops the run at Email Agent, and no email goes out.
• One leg only: each flight line reads flights[0], so the second leg of a connection is left out.
• The request's email is read but not used until you change Send to (step 4).
• Billing: Anthropic bills three Claude calls and each web search, and SerpApi counts two searches a run.
• Open address: once the workflow is published, anyone with the webhook's address can start a run on your keys. Add
  Header Auth in the Webhook node's Authentication if others could find it.
• Your data: the cities, dates and activity wishes go to Anthropic and SerpApi, and the plan goes through Gmail.
• Credentials: the JSON holds none, no personal address and no brand name from its author.