WedLoop
An AI concierge for wedding guests, over plain text message.
- Built by
- Dan Slapelis, solo
- Platform
- SMS and web
- Live since
- March 2026
- Site
- wedloop.io ↗

The problem.
At Dan’s own wedding, every detail was on the website. A custom domain that was easy to remember, a QR code on the save-the-date. Guests still texted the couple.
When people need an answer about an event, they don’t open an app, dig through their inbox, or search for a link. They text someone. Asking guests to download an app or create an account for one weekend asks them to care about the software more than the wedding.
What we built.
A phone number guests can text. WedLoop answers from the couple’s own wedding details, any time of day, in plain language.
The host adds the details
Venue, timeline, dress code, FAQs. WedLoop can import them straight from an existing wedding website.
Guests get a text
Import the guest list and WedLoop introduces itself. No app, no login, no link to lose.
Guests ask, WedLoop answers
Instant, accurate answers from the host’s information, and it remembers the conversation for follow-ups.
Host escalation
Anything WedLoop can’t answer goes straight to the host, who can reply once and teach it for next time.
Smart reminders
Guests get nudged before RSVP deadlines and hotel block cutoffs.
50+ languages
Guests can text in their own language, which matters for destination weddings.
Built for planners
Multiple events and guest lists, managed from one dashboard.
2,000+
Messages sent and received
<10s
Typical response time to a guest’s text
50+
Languages guests can text in
0
Apps guests need to install
When it doesn’t know.
A concierge that guesses is worse than no concierge. When WedLoop can’t answer a guest, it says so and escalates to the host instead of making something up.
The host gets a text with a link to the web app, where unanswered questions wait. They can reply through WedLoop or straight from their own phone, and save the answer so the same question never reaches them again.
The hard part.
Getting a language model to answer a question once is easy. Getting it to answer the same way every time, for every guest, at every wedding, is not.
Most of the engineering went into the agent graph that sits around the model: keeping answers grounded in the host’s details, conversations on topic, and behavior repeatable enough to trust with someone’s wedding day.
Under the hood.
Messaging
Twilio on a dedicated toll-free number, with carrier-approved opt-in and STOP handling built in, as US law requires.
Answering
Frontier language models orchestrated by an agent graph, grounded in each event’s own details.
Guardrails
The same graph keeps conversations on topic and shuts down attempts to misuse the agent.
Running it
Hosted on Google Cloud, with monitoring and alerting that pages us when something breaks.