Vantage
Pre-submission guard X WATCHDOG SPEC · Monad
Vantage simulates your transaction before you sign, scores every contract in real-time, and watches the chain 24/7 — so parallel execution works for you, not against you.
/ 02 · The Thesis
/ 03 · What is actually running
Real simulation, real math, real on-chain data. Exactly one component uses an LLM, and its only job is turning already-computed numbers into a sentence a human can read.
The transaction executes differently than you were shown.
Runs your unsigned transaction against current chain state as a read-only call — no gas, nothing committed — and diffs the result against your quote.
Someone else changed the state while you were reading the page.
Re-checks the contract immediately before you sign. If the slot was claimed or the liquidity drained in between, the simulation reverts and you are told why.
The transaction appears to vanish after you submit it.
Monad's async mempool makes 'still propagating', 'stuck' and 'dropped' look identical to naive polling. This tells them apart.
Execution gets unpredictable when a contract is busy.
Counts recent calls hitting the target contract and scores how contended it is. Above the calibrated threshold, the guard holds and re-checks rather than letting you sign into a moving target.
Unlimited token approvals, granted in one click.
Decodes the calldata, spots an approve() for 2^256-1, and tells you the exact amount you actually needed.
Risk output is dense numbers. That is a communication problem, not a detection one.
Takes the numbers the five modules above already computed and writes one plain-English sentence. It never computes risk and never adds a claim that was not in its input.
Every safety claim needs evidence behind it.
Records every decision and what actually happened afterwards, then feeds that back: contracts get a score, and each contract's hold threshold recalibrates from its own real outcomes.
/ 05 · Gateway
Six deterministic checks. One sentence you can read.
Open the Guard