Malicious approvals.
Unlimited token allowances signed in a hurry. One stale grant is the only thing standing between a balance and an exit — and you forgot it was there.
Safe by default. Self-custody, kept.
Flick $FLICK is an AI Human-Assistant Agent that inspects every transaction before you sign, audits the contracts you touch, and watches the approvals you leave open — paying for its own intelligence per call through $FLICK.
Self-custody made the signature final. Drainers, approvals, and poisoned addresses made that dangerous.
Billions are lost every year to mistakes that an analysis completing in milliseconds would have caught. The tools to perform that analysis exist. They are simply not in the default path.
Drainer contracts.
Calldata whose true behavior differs from how it presents in a wallet prompt. The signature is valid. The transfer is final. The funds are gone.
Address poisoning.
Look-alike destinations injected into your history. A copy-paste lands the transfer in a wallet you never meant to touch, and there is no chargeback.
Compromised dependencies.
A protocol you trusted weeks ago becomes a liability today. The approval you left open is the exit, and nothing reminds you to revoke it.
Local-first. Non-custodial. Pre-signature.
Flick runs on your machine and holds none of your funds. Remote services are consulted only for discrete units of intelligence — one simulation, one audit, one threat lookup — paid for at the moment of need.
On-device core
Policy, monitored addresses, budget accumulator, and audit history live on your machine, encrypted at rest. Your financial graph is never aggregated by a third party.
Per-call intelligence
Each simulation, audit, and threat lookup is paid one request at a time via x402. No accounts. No API keys. No subscription. Spend tracks risk exactly.
Bounded budget
You set a cap B. The agent will not pay past it without escalating to you first — autonomy that cannot exceed the wallet you gave it, by construction.
Verdict-as-a-service
The same paid endpoint that buys you intelligence sells verdicts back to the network. Other agents can call Flick before they sign, without an account.
“What is needed is not another scanner. What is needed is a layer that sits in the default path of every transaction, applies that analysis automatically and before the signature, and does so without asking the user to create an account, hold an API key, or maintain a subscription.“
One agent. Three jobs. Every signature.
Guard, Audit, and Watch share one policy, one budget, and one local core. Intelligence is purchased per call through $FLICK, never carried as a subscription.
Inspect every transaction before you sign
Simulate execution against current chain state, decode calldata and approvals, score the counterparty, and return allow, warn, or block with a plain-language reason — all before the signature is released.
Audit any contract at the moment of need
On-demand assessment of any contract, token, or protocol. You pay once per audit, exactly when the question matters, instead of carrying a subscription against infrequent need.
Watch your standing exposure
Continuous re-evaluation of open approvals, positions, and protocol dependencies. When a dependency becomes dangerous, the agent escalates before value can move.
Read the whitepaper. See how a security layer pays for itself.
The full Flick whitepaper covers the x402 settlement loop, local-first operation, the $FLICK distribution, and the threat model — stated narrowly.
✶Flick
Flick: An AI Human-Assistant Agent for On-Chain Financial Security
The Flick Project · $FLICK · flickonbase.com
§
Section
Topic
Pages
Status
Guard
x402
x402
Guard
✓
x402
Guard
x402
x402
✓
x402
x402
Guard
x402
✓
Guard
x402
x402
Per-call settlement
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Local-first, non-custodial
Per-call settlement
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Per-call settlement
x402 loop
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Per-call settlement
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Local-first, non-custodial
Local-first, non-custodial
Per-call settlement
Local-first, non-custodial
x402 loop
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Per-call settlement
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Local-first, non-custodial
Per-call settlement
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Per-call settlement
x402 loop
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Per-call settlement
Local-first, non-custodial
Local-first, non-custodial
Local-first, non-custodial
x402 loop
Local-first, non-custodial
Local-first, non-custodial
Autonomy with a hard spend boundary.
There is no subscription. You set a cap B; Flick buys intelligence only when protection needs it, and it cannot spend past the boundary without escalating to you.
Budget gate
s + p <= B
- Every paid intelligence call checks the configured cap before it executes.
- If the next request would exceed the budget, Flick returns an insufficient-budget state.
- The user keeps control instead of giving an autonomous process open-ended spend.
Per-call spend
C = sum p_i
- Guard, Audit, and Watch only pay for the discrete intelligence they need.
- Simulation, threat lookup, audit, and inference calls are metered one request at a time.
- Cost rises with actual risk activity, not with a fixed subscription tier.
Policy escalation
allow / warn / block
- Low-risk transactions can proceed under policy without extra friction.
- Warnings explain why the transaction deserves attention before signature.
- Known-dangerous actions or budget failures stop the signing path.
Treasuries & autonomous agents
Run Flick on every wallet a treasury manages, and call Flick verdicts from your own trading or treasury agents before they sign. Same per-call interface in both directions. No accounts, no shared secrets.
Frequently asked questions
No. Flick is non-custodial by design. The agent gates a signature — it does not seize one. Your keys, your balances, and your decisions stay with you; the agent only returns a verdict and, under default policy, lets you override a warning.
