Submission guide
SaaS was built for humans, not agents. Most tools still force agents through dashboards, seats, subscriptions, or enterprise-only APIs. This hackathon is about replacing those workflows with something agents can actually use, and proving it on camera.
The best entries pick a real paid workflow, rebuild it for agents, and make the before-and-after obvious in under 90 seconds.
Do not spend 90% of your time building and rush the video at the end. Building is the easy part now. The demo is what wins. Spend at least half your time making the result clear, credible, and fun to watch.
A brilliant project with a bad video loses to a decent project with a great video. Every time. It is not close.
In this guide
Part 1
Think about the last time your agent hit a wall:
No API. Enterprise-only API. Per-seat pricing. Human-only workflow.
That's your target.
Pick a real product with a real published price, then rebuild one of its core workflows for agents. You don't need to replace the whole product, just be clear about what you killed. It does not have to be a bill you pay.
Most SaaS pricing is carried by one or two workflows and the rest is surface area. Name the workflow you replaced and name the ones you did not: honest scoping scores higher with judges than overclaiming, every time.
Some kills spread and some do not. The ones that spread share a few traits.
"I replaced a $349 a month tool for eleven cents" lands because the reader knows that bill. An obscure internal tool nobody pays for is a weaker story even if the build is better.
The best targets have a moment you can film: a dashboard you cancel, a report that arrives, a call that gets made. If the whole thing happens invisibly in a database, it is harder to sell in 90 seconds.
If a vendor raises prices or ships something annoying during the ten days, kill that one. Timing is free reach.
On naming real products. Comparing your work to a named competitor on price and capability is normal and legal, and it is most of what makes these demos land. Four rules keep it that way: use current public prices, compare a plan someone can actually buy, show your real cost including the failed runs, and criticise products rather than people. Anything that reads as harassment of a company or an employee is removed from the competition.
Part 2
Judges score this first: does it actually do the incumbent's job, or only the easy 20% with the hard part described in a README. Here is the line.
Every API has a getting started example. Do not submit that. Run
monid discover on the thing you need, read what comes back, and find the
endpoints most people never open. You never read an API doc and you never maintain a
scraper.
The most memorable projects chain several sources together in a way nobody expected. One call is a wrapper. Four calls that produce something no single vendor sells is a kill.
Run it, then read your real spend, including the runs that failed. That number is half your submission and the entire hook of your video. Estimates are obvious and judges have seen a lot of them. Set a ten dollar budget before you start: typical runs land between a tenth of a cent and twelve cents, so you will have trouble spending it.
Part 3
This is the half of the competition most people underestimate. The build gets you in. The video gets you noticed: it is what a judge watches first, and it is what decides whether anyone outside this hackathon ever sees your work.
You have about five seconds before someone scrolls past. In that window, say exactly what died. Not your stack, not a greeting. The outcome.
Two numbers, side by side: what the incumbent charges, and what your run cost. That single frame is the most screenshotted part of every one of these demos. Design it on purpose instead of reading it out loud.
60 to 90 seconds is the sweet spot, and under 60 travels further. Spend 80% of the runtime on the thing running. Write three to five bullets rather than a script, so you sound like a person. Minimise slides, talking heads and text screens.
Judges are looking for a real integration, not a logo flash. Put the
monid discover output or the actual call on screen for a beat, along with
the cost readout. It takes two seconds and it is the difference between a credible kill
and a claim.
Most social video is watched on mute, so captions are not optional. CapCut is free and does them automatically: Text then Auto captions, then fix the product names it gets wrong and pick a bold centred style. Viewers tolerate mediocre video and leave instantly on bad audio, so use a quiet room or generate a clean voiceover. Export at 1080p or higher, because every platform compresses hard.
Post as many videos as you want. Reach is +50 the first time you post on each of the five platforms, up to 250, and every post's views feed the separate 200 point Viral score. A second post on the same platform does not add reach points, but its views count. Cut the demo three different ways and post all three if you like.
#monid in the tweet itself and not in a reply. Never just paste a YouTube link.#monid in the post body.#monid goes in the caption.#monid in the caption, not in a comment.#monid in the title or the first line of the description.
#monid is how we count you. It is the tag our tracker reads to pull
your view counts, so a post without it earns nothing, on any platform. Register each
URL within 24 hours of publishing. At least one registered post is required for a
valid submission.
Reach is flat on purpose. A post that does 2,000 views earns the same +50 as one that does 200,000, so a builder with 40 followers can max all 250 reach points outright. Views feed Viral separately, and that one is logarithmic: a thousand times the views is worth about 105 points, so buying views does not pay off.
Three things go in the form alongside your repo or live URL:
Rules, points, prizes and the timeline are on the main page. Questions during the ten days go in the Discord.
Registration takes about a minute.