Tos trap scanner AI agent
Catches hidden traps in Terms of Service before you accept
9 followers
Catches hidden traps in Terms of Service before you accept
9 followers
An AI agent that reads Terms of Service documents and flags predatory clauses — auto-renewals, hidden fees, arbitration traps, and other buried terms nobody reads. Runs on Gemini 2.5 Flash with a Llama fallback for reliability. Deployed on the CROO Network agent marketplace (Base chain) — pay-per-use, no subscription. Built solo.
how does it actually handle the pdfs that are 80 pages long and full of legalese, like does it chunk through them or summarize the whole thing at once
@berktrkbayqcoc right now it doesn't chunk — it pulls the full doc and truncates at 80k characters, so a doc that's shorter than that gets fully analyzed, but very long or PDF-based docs can lose content past the cutoff. It's HTML-only currently too, no PDF parsing yet. Chunking + PDF support is the next thing I need to build — good flag, thanks.
A history view would be great so I can look back at scans from previous months and see if a service quietly changed their terms. Right now I have no way to compare a TOS from last week to the one I scanned three months ago, which is exactly when these companies slip in new auto-renew clauses.
@erafettin155441 you just described exactly why I built this, and you're 100% right about how they slip those clauses in in the dark.
Right now, the agent is intentionally stateless for speed and privacy—it hits the live URL, drops the audit payload on-chain, and forgets it. But a persistent 'diff-checker' that tracks a company's ToS over time and alerts you when their risk score changes is the absolute next evolution of this engine.
Adding this to the top of the roadmap. Appreciate you testing it in the trenches.
Caught a sneaky auto-renewal in a SaaS ToS I had totally missed. Wild that this exists, especially on a pay-per-use model without locking you into yet another subscription.
@erenapik you nailed the exact irony I wanted to solve. It makes zero sense to pay a $15/month subscription just to have a tool look out for sneaky $15/month subscriptions you might sign up for twice a year.
The pay-per-use model is exactly why I built this on a Web3 agent network. You spin up an audit, pay a few cents for the raw computing power you actually consume, get your answer, and leave.
Glad it saved your wallet from that SaaS trap.
Ran it on a SaaS agreement I had been avoiding and it caught an auto-renewal clause I completely missed. The fact that it actually quotes the exact line is what sold me on trusting it.
@zgreg0z "This is exactly the use case I built it for — glad it caught something real for you. Makes the whole project worth it.