The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
A multi-agent AI system I built to run the day-to-day operations of my own SaaS product. Specialized agents handle development, QA, customer support and customer success, so a live product with ~500 users runs with a team of one. Built on OpenAI and Anthropic Claude models with TypeScript, Python and Supabase/Postgres. I build the same kind of agents for clients: support bots, internal assistants, automated QA and workflow automation.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Proper Form — Social Media Branding System
A flexible social media system created for Proper Form as part of the wider brand identity.
The template system was designed to adapt across different post formats, banners and platforms while maintaining a consistent visual language.
Rather than repeating one rigid layout, the system gives the brand room to vary composition, imagery and content while remaining clearly recognizable.
Created by Lepisov Branding in collaboration with Creative Mints.
Logo design: Ilya Gorchanyuk
Creative direction: Dmitry Lepisov
Brand identity + social system: Lepisov Branding × Creative Mints
This is how a social system should work: strong enough rules that the brand reads at thumbnail size, loose enough that the feed never looks copy-pasted. Do you hand clients a Figma kit for this, or build it in something their team edits themselves?
FirstReply — reply first, win the booking
Description
Event venues lose bookings because they reply a day late. By then the client has booked somewhere else. FirstReply answers every inquiry with a priced proposal and a held date in seconds.
The automation
1. An inquiry arrives from the venue page, or staff paste in an email.
2. AI extracts event type, date, guests, budget and requests, each with a confidence level and the exact words it came from.
3. Missing date or guest count? It writes a clarifying reply instead.
4. It matches the smallest free space that fits (or the two nearest free dates when it's booked), prices the right package and holds the date for 5 days.
5. The client gets their own proposal page and books with one tap. The hold becomes a confirmed booking.
6. Unsure details, over-budget totals or Review mode send it to a review queue first. Humans step in only when needed.
7. Views, acceptances and expiring holds update the pipeline and alert the team live. Every step is logged to the second.
The design A full design system first (tokens, type scale, components), built as a clickable prototype, then rebuilt in Base44 from prompts. Client pages are mobile-first; the venue side is desktop-first.
Try it Client side: firstreply.base44.app · Venue side: firstreply.base44.app/dashboard
Vote on the design direction for the hero section.
Tula is an AI therapist app that helps boost relationships, reduce anxiety, and improve life using language patterns.
I am building both the AI and the UI
3 voted
30%
7 voted
70%
10 votes
Closed
Voted Human. The photo feels warmer and fits a therapist app better than the waveform.