Connect Botpress to Google Calendar through the official integration, then build a flow that collects appointment details, uses Check Availability, asks the visitor to confirm a returned slot and runs Create Event. Test permissions, time zones and failures before publishing; creating an event alone does not prove that invitations or conflict protection work.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThe manageable first project is one appointment type on one staff calendar, with a fixed duration and clear office hours. Avoid starting with technician routing, equipment capacity or automatic rescheduling. A calendar connection supplies actions; the business must still define when the bot may use them and what evidence makes a booking successful.
Define one appointment the calendar can represent accurately
An illustrative car repair garage wants customers to book a 30-minute estimate discussion with its service adviser. The booking reserves the adviser's time, not a workshop bay or a repair. The bot should state that distinction in the event description and customer confirmation. All times, counts and internal costs in the examples are teaching assumptions rather than observed business results.
Use a test calendar for the first build and a proposed booking window of 09:00 to 12:00. With 30-minute appointments there are six candidate intervals. Add an illustrative busy event from 10:00 to 11:00; four candidates remain: 09:00, 09:30, 11:00 and 11:30. Those expected results give the owner an independent way to test the integration's availability output.
Record the appointment title, duration, allowed days, booking window, time zone and confirmation route in a small worksheet. Give calendar ownership to the adviser or account administrator. The owner should also decide how a customer cancels and how staff see new bookings. A created event that nobody checks is not a useful operational handover.
A cleaning company could instead reserve a 20-minute estimating call, while a furniture maker might reserve a 45-minute design discussion. Each needs its own capacity rules. Neither example should imply that a call reserves cleaning staff or manufacturing time. If a service requires several resources simultaneously, a single calendar free-slot check is insufficient evidence that the business can fulfil it.
How do you connect Botpress to Google Calendar?
To connect Botpress to Google Calendar, install the official Google Calendar integration from Explore Hub in Botpress Studio, then authorize your Google account. The official Google Calendar integration guide says you need a published bot and a Google account for basic configuration. In Botpress Studio, select Explore Hub, find Google Calendar and choose Install Integration. In Configuration, select Authorize Google Calendar and follow the account connection prompts.
Confirm the Google account before accepting access. If several accounts are signed into the browser, identify the one that owns or can edit the intended test calendar. Then perform a harmless test event on that calendar and inspect it directly. Do not use the apparent success of the authorisation screen as the only evidence that the bot can write where you expect.
The guide also documents manual configuration using a service account, a credential used by software rather than a person signing in interactively. That path runs through a Google Cloud project: enable the Google Calendar API, create the service account and its JSON key, share the calendar with the account's client email at Make changes to events, then choose Configure manually in Botpress and fill in Calendar ID, Client Email, Private Key and, for Workspace, Impersonate Email. It is a reasonable handover to an administrator if those terms are unfamiliar; the basic authorisation route should be the owner's first experiment.
Keep credentials out of prompts, knowledge documents and customer messages. If you do take the manual path, treat the downloaded JSON key like a password: the guide warns that it contains sensitive credentials. Paste its private key only into the Private Key field, delete stray copies from the downloads folder and shared drives, and share just the one booking calendar with the service account, not every calendar the business owns.
Use the beginner Botpress tutorial if Studio's nodes and cards are unfamiliar. A node is a conversation step, and a card performs a task inside it. For booking, prefer an explicit sequence of steps so a visitor's confirmation and the calendar write are easy to trace.
Does the Botpress Google Calendar integration work with personal Google accounts?
Botpress's Google Calendar integration manages events on Workspace and personal accounts, but its guide lists Meet links and attendee invitations for Workspace only. The guide lists event creation, reading, updating and deletion for both Google Workspace and personal accounts. Its limitation table lists Google Meet link creation for Workspace, but not personal accounts, and attendee invitations for Workspace with domain-wide delegation. These are the integration's documented boundaries; they are not a statement that personal Google Calendar users can never invite anyone manually.
Domain-wide delegation allows a configured service account to act for a user within a Workspace setup. The guide requires it for the manual service-account path to create Meet links, send invitations or act on users' behalf, and warns that without it the service account can only manage its own calendar. It also describes an Impersonate Email setting, which must name a real user in the Workspace domain; personal Gmail accounts support neither delegation nor impersonation. Have the administrator review this path rather than treating a missing invitation as a reason to expose or repeatedly replace a private key.
| Expected result | What to establish | Customer wording until verified |
|---|---|---|
| Event in staff calendar | Write permission and correct destination | State success only after Create Event succeeds |
| Customer invitation | Supported account configuration and delivery test | Do not promise an email invitation |
| Meet joining link | Supported Workspace configuration and returned link | Do not invent a video link |
| Conflict-safe booking | Availability checks plus coordination of writes | Offer a request if conflict handling is incomplete |
For the garage's first test, omit video links entirely. If the configured account cannot send an invitation, make the confirmation route explicit before customer launch. The owner could use a separate, deliberately tested notification process, or choose a scheduling-page handoff instead. The Botpress and Calendly tutorial describes that simpler booking-page route.
How should a Botpress bot collect a booking request and show slots?
A Botpress booking bot should identify the appointment type, collect the requested day, establish the visitor's time zone, call Check Availability for the allowed business window and show only the slots that come back. Use Capture Information cards for structured answers and workflow variables for values needed later. A variable is simply a named piece of information that another step can read. Capture Information has a Datetime type that turns phrases such as “next week” into a date object, which is handy, but read the result back to the visitor before using it.
For the first trial, use a small list of approved dates rather than asking AI to interpret every possible phrase. A visitor saying “next Friday afternoon” may mean a different week or a time outside office hours. Ask a clarifying question and display the resulting full date. Do not let a plausible interpretation become an appointment without the customer's explicit confirmation.
Written out as nodes, the garage's test bot looks like this. Each line is one step you can test on its own in the Emulator, and the numbered-choice approach avoids depending on the bot to read a free-text time correctly.
Illustrative node map, garage estimate bot (test calendar):
1. Capture Information, Single Choice: appointment type ("Estimate discussion")
2. Capture Information, Single Choice: date from an approved list
3. Check Availability: 09:00-12:00 with offset, Slot Duration 30, Timezone set
4. No free slots: offer the next approved date or the enquiry form
5. Text: list returned times as 1-4, with the full date written out
6. Capture Information, Number: the visitor's choice, 1 to 4
7. Capture Information, Confirmation: "Reserve ... ?" yes or no
8. Check Availability again for the chosen half-hour; stop if busy
9. Create Event: Summary, Description, Start and End from the chosen free slot
10. Text: confirmed date, times and zone; store the Event ID
The official integration's action definitions expose Check Availability inputs named Start Time, End Time, Slot Duration (minutes) and Timezone. Slot Duration accepts 15 to 480 minutes; set it explicitly to 30 for the garage example rather than relying on the default. Timezone only controls how the readable labels are written. The Start and End values decide which times are checked, so each must include the date and the correct offset for your business, in the ISO format the field describes, such as 2027-01-12T09:00:00+02:00 for a business running two hours ahead of UTC.
Illustrative Check Availability values for a test only:
Start Time: 2027-01-12T09:00:00Z
End Time: 2027-01-12T12:00:00Z
Slot Duration (minutes): 30
Timezone: UTC
Expected with a busy event at 10:00–11:00 UTC:
09:00–09:30, 09:30–10:00, 11:00–11:30, 11:30–12:00.
Illustrative Formatted Free Slots output for that case:
9:00 AM – 9:30 AM
9:30 AM – 10:00 AM
11:00 AM – 11:30 AM
11:30 AM – 12:00 PM
Replace the sample date and UTC with your approved dates and time zone.
The date is invented sample input, not a vendor release date. UTC is used to make the example unambiguous; it is not a recommendation to run every business in that zone. If your operating zone changes its clock seasonally, put the offset that applies on the chosen date into Start and End, and the matching named zone into Timezone so the labels agree. A fixed offset copied from today's test can be an hour wrong for an appointment after the clocks change.
The action returns machine-readable freeSlots, readable formattedFreeSlots and busy slots. In the integration's code, the free slots come back as UTC timestamps ending in Z, while the readable labels use a 12-hour clock and show times only, with no date. So ask about one day at a time, or put the date into the message yourself: a list reading “11:00 AM – 11:30 AM” across a three-day search gives the visitor no way to tell which day it means. Present only returned choices and keep the original start and end values for the chosen slot; never turn a displayed “11:00 AM” back into a newly guessed timestamp.
If no slots are returned, send a useful message and offer another permitted date or a staff enquiry route. A pest-control firm's illustrative inspection discussion may have no remaining afternoon slots; the bot can offer the next approved discussion window. It must not create an evening appointment outside the agreed hours just because the calendar appears empty then.
If you want help fitting a booking bot into how your business actually runs, I offer one-to-one AI consultations for small businesses.
How do you stop a Botpress booking bot offering slots outside office hours?
Restrict the Check Availability request to the permitted booking window, because the integration returns free intervals for whatever range it is given and has no working-hours rule of its own. The integration's published availability implementation asks Google for the busy times of the one calendar configured in the integration, builds back-to-back intervals starting exactly at the requested Start Time, and removes every interval that overlaps a busy period. It is not a complete scheduling policy engine. Supplying the whole day therefore asks a different question from supplying the garage's 09:00–12:00 booking window.
For example, a 30-minute slot grid starting at 09:00 with a busy event from 09:20 to 09:50 should lose both 09:00–09:30 and 09:30–10:00. Each interval overlaps the existing event. This is a concrete boundary test for the owner: a partial overlap matters, not merely whether another event starts at exactly the same time.
Buffers need their own policy. If a 20-minute cleaning estimate call needs ten minutes of preparation afterwards, reserve a 30-minute block in the calendar or implement equivalent buffer handling and test it. Checking a 30-minute free slot but writing an event lasting only 20 minutes can leave the intended buffer available to another process. The displayed customer call duration and reserved staff interval must be explained consistently.
Do not assume this one integration checks every diary relevant to a job. An adviser may block personal commitments elsewhere, or a garage appointment may also need a bay. Consolidate the necessary busy information into the approved scheduling arrangement or use a system that can coordinate those resources. A blank interval on one calendar is only evidence about that calendar.
Why confirm with the visitor before Botpress creates the event?
Ask the visitor to confirm the service, full date, start, end and time zone with a yes-or-no question before the Create Event card runs. A suggested message is, “Reserve a garage estimate discussion on 12 January 2027, 11:00–11:30 UTC? This reserves adviser time, not a repair.” Use the visitor's actual selected details in production.
The calendar-write step should be reachable only through an explicit yes response. A request to see another day must return to availability, and a no response must not create anything. Use a deterministic confirmation branch, meaning a defined route for each answer, rather than asking a conversational model to decide whether the visitor “sounds ready”. Capture Information's Confirmation type does this job: it stores the answer as true or false in a variable, and a transition on that variable leads to Create Event only when it is true.
Map the chosen interval into Create Event. The official Event schema names Summary, Description, Location, Start Date/Time and End Date/Time, plus Attendees, Send Notifications (on by default) and Enable Google Meet (off by default). Its description of Attendees says email invitations are sent automatically, but the guide's table lists attendee invitations as unsupported on personal Gmail accounts and dependent on delegation in Workspace, so on a personal account leave Attendees empty and confirm in chat instead. For the garage, use a summary such as “Estimate discussion” and a description stating the approved purpose. Keep sensitive customer problem details out of the event title because staff may see titles in several shared views.
The start and end should come from the selected free-slot values, not from newly generated prose. Store the returned created-event identifier in the booking record your implementation deliberately maintains. Then show the customer the confirmed details only after the action succeeds. If you cannot confidently map these fields and inspect their results, ask a helper to complete and demonstrate that specific connection.
Google's event-creation guide says start and end are the only required fields; attendees receive an email only when the request asks Google to send updates, and Meet details need conference data switched on separately. Test the invitation independently if your setup supports it. An event in the staff calendar, an invitation sent and an invitation accepted are three different observations; avoid collapsing them into one cheerful “all done”.
How do you prevent double booking in a Botpress appointment bot?
Recheck the selected interval immediately before the Create Event card runs and stop if another appointment has taken it, because availability changes while a visitor is choosing. In an illustrative two-tab test, the first visitor sees 11:00, another appointment occupies it, and the first visitor then confirms. The correct response is to offer fresh choices, not to honour an outdated display.
A recheck reduces the stale-slot problem but does not make the read and write atomic. Atomic means no other booking can slip between the availability check and reservation. The official Create Event implementation delegates to event creation; it does not document a lock that coordinates every writer. Do not advertise guaranteed double-booking prevention from those two cards alone.
If simultaneous website bookings are possible, require a booking mechanism that coordinates competing requests, or keep the bot at appointment-request status until staff approve. Get help for this boundary and ask for a test where two visitors confirm the same slot together. One should succeed and the other receive a genuine alternative under the implemented policy; two overlapping confirmed events are a launch failure.
Retry handling is equally important. Suppose the event is created but the bot loses the response. Blindly running Create Event again may create a duplicate. Preserve a request reference and reconcile against the calendar before repeating a write. The action definitions also provide List Events, which returns up to 100 events from a chosen date by default, plus Update Event and Delete Event, which take the Event ID; that is why the identifier from Create Event must be stored. The correct reconciliation method still depends on your implementation and must be demonstrated rather than guessed.
A customer pressing confirm twice is another test. The desired result is one booking, one event identifier and one clear confirmation. If a basic no-code flow cannot establish that behaviour, use a scheduling-page handoff or an approval queue for the public version. This limitation is a practical reason to buy a small piece of expert help, not to abandon the owner's useful work on appointment rules.
When is a Botpress booking bot ready to publish?
A Botpress booking bot is ready to publish when every planned test journey gives the expected outcome and the bot, the calendar and any confirmation message all agree on the booking. Set each expected outcome before the test runs.
Use an illustrative acceptance set of 12 journeys: three ordinary appointments, two busy-slot requests, one no-slots day, two time-zone checks, one customer declining confirmation, one duplicate click, one simultaneous-slot attempt and one uncertain response after creation. Set the expected outcome for each before running it. Keep test data separate from customer appointments and remove test reservations deliberately afterwards.
For the main numerical example, confirm that the 10:00–11:00 busy event leaves four candidates and that booking 11:00–11:30 leaves three on a fresh availability check. Inspect the calendar directly, not only the bot's text. If the action says success but the wrong calendar received the event, the customer-facing journey has failed even if no technical error appears.
For the zone test, compare two views of the same instant. An illustrative 11:00 UTC start appears as 13:00 at a fixed +02:00 offset. Both must refer to the same half-hour appointment. Also test a date around a clock change if your actual chosen zone has one. Display enough date and zone information that a visitor can detect a mismatch before confirming.
Botpress usage is a separate operating cost. The pricing page lists Free with 100 included conversations without a renewal period beside the figure. Plus is shown at $150 a month billed annually with 250 conversations per month and $25 of AI usage. Check your own account before testing; current pricing wording includes emulator conversations, so do not assume all trial journeys are quota-free.
Budget owner time as well. A suggested first pass allows 30 minutes for appointment rules, 45 for account connection and field mapping, and 90 for the acceptance journeys and repairs. These are planning allowances, not measured completion times. Obtain separate quotes for any Workspace administration or conflict-handling work rather than assuming the bot subscription buys that implementation.
When the test bot passes, publish the intended version and repeat the essential journeys through the actual website Webchat. Keep the ordinary booking or enquiry route visible outside the widget. If chat is silent, use the Botpress response-failure checks; if the calendar action fails, retain its error and the selected interval for the maintainer.
Finish the handover with the destination calendar, account type, approved window, duration, zone, confirmation route, cancellation procedure and recovery contact. Ask a colleague to find a test booking using its saved event reference. Use the Botpress expert-scoping guide if the remaining work involves permissions, coordination or retries. The public promise should match the behaviour you can demonstrate under those failures.
If you keep your acceptance-test results in a spreadsheet, try The Complete Excel Course with AI Integration on Coding Liquids.
Further reads
- Botpress Pricing Plans: Free, Plus and Team for a Small Business — Understand conversation and AI usage costs.
- Botpress for WordPress: Embed the Webchat and Test It First — Check the website installation after the workflow passes.
- Botpress for WhatsApp: What a Non-Developer Can Set Up Alone — Assess another channel separately from calendar logic.
Browse all AI-for-business tutorials
Sources: Botpress Google Calendar integration guide; the integration's action definitions, Event schema, availability and event-creation code in Botpress's public GitHub repository; Botpress Capture Information documentation; Botpress pricing page; Google Calendar API create-events guide.
