Start here
Set up your workspace.
Connecting the connector gets your AI talking to Presso Events. It doesn't make your AI think of Presso Events. That takes a bit of one-time setup: a place for your AI to work from, and a short set of instructions telling it what to do there. Do this once, before your first event, and every event after it gets easier.
The default: one claude.ai Project per organiser
Create a single claude.ai Project for your organisation - not one per event. Name it after your organisation, not the event. Events come and go inside it; the Project is the working relationship.
Three reasons this is worth doing, not just a nice-to-have:
- It survives every new chat. A Project's instructions and files sit server-side. Start a fresh conversation next week and your AI still knows your folder convention, your organisation name, and how you like things done. A one-off chat forgets all of it the moment you close the tab.
- It works from your phone. You don't need a laptop to check on an event or fix a typo in the agenda. Open the claude.ai app, open the Project, ask.
- The Google Drive connector lives there. If your event files live in Drive rather than on a laptop, the Drive connector is attached at the Project level, so every chat inside the Project can read your folder.
This needs a paid claude.ai plan - Pro, Max, Team or Enterprise. Adding a custom connector (the Presso connector) is a paid-plan feature. On the free plan, adding the connector doesn't fail loudly - it just does nothing, and that reads as a Presso bug when it isn't one. Check your plan before you start.
If you're technical and prefer working from a terminal, the same convention works with a local folder plus Claude Code instead of a claude.ai Project: same folder layout, same idea, files in git. Nothing below changes - just swap "Project" for "repo".
The folder convention
Whether your files live in Google Drive or on disk, use the same layout. Everything is Markdown - a plain text file with light formatting - because you'll edit these from your phone and your AI reads them directly. Not YAML, not CSV: those are formats for parsers, and nothing here parses your files. Your AI just reads them, the way you would.
Presso - <Organiser Name>/
brief.md dates, venue, capacity, what the event is
tickets.md types, prices, quantities, sale windows
agenda.md sessions: time, title, room, speaker
speakers/
ada-lovelace.md name, title, company, bio
ada-lovelace.jpg
assets/logo.png
In speakers/, a speaker's write-up and their photo share a filename stem: ada-lovelace.md pairs with ada-lovelace.jpg. That's the only rule - it's how your AI knows which photo belongs to which speaker when it reads the folder.
What this actually is - and isn't
This is the single most misunderstood thing here, so it's worth being blunt: Presso cannot read your folder. We hold no file access and no Google Drive credentials. Nothing on our side ever opens brief.md or looks inside speakers/.
What happens instead is that your AI reads the folder, because it's sitting right there in the same Project, and then calls Presso's tools with the values it found - the ticket price from tickets.md, the session times from agenda.md. The folder is a convention your AI follows, not a system Presso is plugged into.
That's why nothing syncs by itself. Edit tickets.md and the live event doesn't change until you ask your AI to read the file and update the tickets. This is deliberate: it's also why your AI never invents a price on its own - it either reads one from your folder or asks you.
Paste this into your Project's custom instructions
This is the block that actually does the work. Paste it into your Project's custom instructions once, and every chat in the Project inherits it. Replace <Organiser Name> with your organisation's name - that's the only thing you need to change.
I organise events under "<Organiser Name>" on Presso Events, using the Presso MCP connector.
For anything about an event - dates, venue, tickets, agenda, speakers, publishing, attendees - use the Presso connector's tools rather than answering from memory or guessing.
Before asking me a question, check whether my workspace folder already answers it. My files follow this layout:
Presso - <Organiser Name>/
brief.md dates, venue, capacity, what the event is
tickets.md types, prices, quantities, sale windows
agenda.md sessions: time, title, room, speaker
speakers/
<speaker-name>.md name, title, company, bio
<speaker-name>.jpg their photo, same filename stem as the .md
assets/logo.png
Read the relevant file before asking me something it already contains.
Never invent a ticket price, date, or capacity number. If it isn't in my folder and I haven't told you, ask me - don't guess, and don't carry a number over from a different event.
A speaker photo can't be sent through this chat. When a photo is needed and isn't already in my folder, ask Presso for an upload link, tell me it expires in 30 minutes, and remind me to open it on my phone and send the photo there. Once uploaded, the photo becomes part of the public event page - it isn't private.
Publishing a paid event requires me to complete Stripe onboarding. If publishing returns a Stripe link, give it to me plainly and tell me to complete it in a browser, then come back. A free event publishes without this step.
Presso deliberately refuses to publish until it has read back my full, confirmed ticket list. That's a safety check, not a bug - it stops a wrong price going live by mistake. If it happens, read tickets.md (or ask me) and confirm the prices before trying again.
A freshness recipe
Your AI doesn't run anything on its own schedule - it only acts when you talk to it. A stateless connector like Presso's can't reach into your calendar and check things while you're not looking. So instead of a setting you switch on, here's a recipe: a prompt you save and run yourself, on whatever cadence you like, in whichever client you use to schedule things.
Three worth having:
Read agenda.md in my workspace folder and compare it against the live agenda for <Event Name> using get_agenda. Tell me anything that's different, then update the live agenda to match my file.
A Monday agenda check. Catches the case where you edited the file but forgot to tell your AI, or where you updated the live event by hand and your file drifted.
Read tickets.md in my workspace folder and compare it against the live ticket types for <Event Name> using ticket_types. Tell me any differences in price or quantity before changing anything.
A weekly ticket check. Reports differences rather than changing prices on its own - worth running before a Friday send to your list.
Give me the status of <Event Name> using list_events with that event: is it published, how many tickets have been issued, and how many people have checked in.
A pre-event and day-of status pull. Useful daily in the week before doors open, and every couple of hours on the day itself.