Anthropic AI Hackathon · lablab.ai · vøiddo studio
understand any contract
before you sign.
Deal Clarity Agent sends your contract to Claude and returns a structured risk report — obligations, traps, missing clauses, and a one-sentence verdict. Every finding is tied to an exact quote from the document.
choose a sample contract to analyze
one-liner verdict
loading…
obligations
what each party must do.
risks
clauses that could hurt you.
missing elements
what should be there but isn't.
architecture
how Deal Clarity Agent works.
contract input
User submits contract text (paste, PDF extract, or plain text). No PII is stored.
Claude analysis
Contract sent to Claude with a structured JSON schema prompt. Claude returns obligations, risks, missing elements, and a plain-English verdict.
structured report
Every finding is validated against the schema, severity-scored, and linked back to a direct quote from the contract.
plain-English output
Results rendered as a readable risk dashboard — no legalese, no ambiguity.
real API call (from deal_clarity.py)
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-opus-4-7",
max_tokens=4096,
system=SYSTEM_PROMPT, # structured JSON schema instruction
messages=[{
"role": "user",
"content": f"Review this contract:\n\n{contract_text}"
}]
)
report = json.loads(message.content[0].text) # guaranteed valid JSON
the problem
contracts are written to protect the drafter, not the reader.
lawyers charge $400/hr
A solo founder or freelancer can't afford a proper contract review for every NDA or service agreement they sign.
traps are buried in boilerplate
IP assignment clauses, perpetual non-competes, and uncapped liability live in paragraphs nobody reads until it's too late.
Claude reads what humans skim
Every clause, every obligation, every missing element — surfaced in plain English with the exact contract quote as proof.