Sep 1 to Sep 10, 2026 · 10 days
SaaS was built for humans, not agents. Agents still can't use most tools without expensive subscriptions. Join us to kill the old SaaS model and build the next generation of products for agents.
Total prize pool
$4,015
Claude and Monid credits, split evenly across the top three.
Winners announced
Sep 15, 2026
Judges 400, reach 250, viral 200. Highest total out of 850 takes it.
Why they have to go
Built for clicks, not calls. Your agent has to use a browser, or give up.
The API exists, but only behind an enterprise plan or a sales call.
Agents don't need seats. They need a few calls, not another dashboard.
SSO, MFA, manual exports: workflows made for people, not agents.
That's what we're here to kill.
How it works
Not the subscription you resent. The thing your agent could not do, sold to humans at a public price.
How to choose one → Step 02One Monid key, any framework, every call priced on its own. It has to run on live data.
What counts as a kill → Step 03A video under 90 seconds with their price next to your real cost, posted everywhere.
How to shoot and post it →Tue Sep 1
Thu Sep 10
Tue Sep 15
What is Monid
They charge per seat, per month. Monid charges per call. One key, no plan, no minimum, no sales call. Chain four of these and you have an output no single vendor sells.
Give this to your agent
$ set up https://monid.ai/SKILL.md
and let it take it from there.
Monid
Google Reviewsscraping
Apifyweb scraping
Browserbaseweb automation
People Data Labspeople search
Wokeloprivate markets
Suzanne3D modeling
Exaweb search
Saperlyagent phone
Gloriareal-time newsPrizes and scoring
Prizes
$4,015 total1st place
$2,000
$1,000 Claude credits, $1,000 Monid credits
A case study and a feature video from the Monid account
2nd place
$1,200
$600 Claude credits, $600 Monid credits
A feature in the winners thread
3rd place
$800
$400 Claude credits, $400 Monid credits
A feature in the winners thread
Worst idea that worked
$15
$15 in Monid credits
The same case study as 1st place
Worst Idea That Worked is a judges' pick. It ignores the score completely, so an entry that places nowhere can still take it. It does still have to run on real data.
Scoring
Judges
400
Four criteria, 100 each
Reach
250
Five platforms, 50 each
Viral
200
Total views, log scaled
Final score
850
Highest three take 1st, 2nd, 3rd
400 points
Every judge scores each criterion out of 100. Their four totals are averaged, so 400 is the scale, not a placement bonus.
250 points
Platform coverage, not post count. Post as many videos as you want: the first one on each platform is +50, the rest add views.
| Platform | Points |
|---|---|
| X | +50 |
| +50 | |
| +50 | |
| TikTok | +50 |
| YouTube | +50 |
| All five | +250 |
Flat, regardless of performance. A post at 2,000 views earns the same +50 as one at 200,000, so a builder with 40 followers can max all 250 of these outright. At least one registered post is required for a valid submission.
200 points
The one place views count. Your total across every registered post, on a log curve against whoever leads.
# locked before we open Viral = 200 × log10(1+V) ⁄ log10(1+Vmax) V your total views, every post Vmax the leader's V
If the leader ends on 500,000 views
| Your views | vs leader | Viral |
|---|---|---|
| 500,000 | 1× | 200 |
| 50,000 | 1/10 | 165 |
| 5,000 | 1/100 | 130 |
| 500 | 1/1000 | 95 |
| 50 | 1/10000 | 60 |
A thousand times the views is worth 105 points, which is 12% of the 850 total. Big accounts have an edge and it is not a steamroller. Bot views on X run about seven cents per thousand, so buying your way up a log curve costs more than it returns, by design.
Judges
Do not start from your card statement. Start from the last time your agent could not do something: no API, an API behind a sales call, or pricing that assumes a human with a seat. That wall is your target.
Somebody sells the human version of that job, and their price is on a public pricing page. That published price is the only hard requirement, because the comparison is the entire point. There is no approved list, no category restriction, and no advantage to picking something famous.
Your build has to do the job a paying customer buys that product for, end to end, for at least one real use case. Running on live data, in front of a judge, handling bad inputs and empty results. A UI with mock data behind it is not an entry.
You do not have to match every feature. Name the workflow you replaced and name the ones you did not, because honest scoping scores higher than overclaiming every time.
Teams up to three. Register one entry with one bound social handle, name the whole team in the submission, and split the prize yourselves.
The Monid integration and the kill have to be new work built during the ten days, and your commit history has to show it. Building on top of something you already had is fine as long as the judged part was written inside the window.
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 failed runs, and criticise products rather than people.
Anything that reads as harassment of a company or an employee is removed from the competition.
You own it entirely. By entering you give Monid permission to show your submission, video and name in coverage of the hackathon. No license to your code, no assignment, no right of first refusal.
Sep 1 to Sep 10. We send your Monid key the morning it opens.