Behavioral Attestation
& Intent Detection

Detect transaction intent. Prove every action.

→A post-trained transformer model detects intent and checks every action.
→A cryptographic receipt makes every action independently verifiable.
→Every verified action compounds into a tamper-proof agent reputation.

Learn more

zeroproof / intent → proof

tool callintent modelproof

send_money→aligned→sha:7f3averified

buy_item→aligned→sha:9d2cverified

checkout→aligned→sha:4e1bverified

wire_funds→drift→sha:2a8frejected

pay_invoice→checking→sha:6c5epending

issue_refund→aligned→sha:1d9averified

fund_wallet→aligned→sha:8b2dverified

transfer_out→drift→sha:3e9crejected

renew_plan→aligned→sha:5a1fverified

split_bill→checking→sha:0c7epending

charge_card→aligned→sha:2f6bverified

Products

Intent models. Action proofs.

01 / Behavioral Attestation

Prove every action an AI agent takes

After an agent acts, no verifiable record exists. ZeroProof issues a receipt for every tool call, signed by both the agent and the tool server, anchored on-chain.

# example receipt

tool: web_search

agent: gpt-agent-7f3a

proof: sha256:a1b2c3…

chain: base:0x9d2c…

02 / Intent Detection

Prove the agent did what was asked

An agent can call the wrong tool and still sound certain. ZeroProof's first model is a small transformer, post-trained for transaction intent. It flags tool calls that drift from user intent before they run.

// model 0 · balanced eval, n=4,000

Intent routing (macro)

0.70

Valid output

1.00

Detail extraction

0.61

base gemma-3-1b routes at 0.14 macro

View on Hugging Facev0

Research

From the lab

We fine-tune small, task-specialized models to read what a person actually means before an agent acts on it. Intent Model 0 is our first, focused on payment and commerce conversations.

Intent Model 0 on Hugging Face ZeroProof on Hugging Face

Protocol

Verified execution flow

ZeroProof is a model lab building the trust layer for AI agents. It post-trains models to read user intent, then proves every action with a verifiable receipt.

Agent issues a tool call

An MCP-connected agent sends a tool request. ZeroProof intercepts it before execution.

tool_call →

Intent model runs

Before it runs, the intent model checks the call against stated user intent.

intent: aligned

Proof is generated

If intent is confirmed, a bidirectional cryptographic receipt is issued, signed by both the agent and the tool server.

proof: sha256:7f3a

Reputation is updated

The verified action is anchored on-chain. Each receipt contributes to a tamper-proof, auditable agent reputation.

reputation +1

From “trust me” to “verify it yourself”

ZeroProof is in early access. Get notified when the intent model and Python SDK ship.

Signed Receipts

Intent Guardrails

Agent Reputation

Audit Trail

Behavioral Attestation& Intent Detection