Limits

Hard refusals

These are not configurable.

• It will not hold a deploy key. Mainnet transactions are prepared and handed to you. There is no mode where a deploy happens without a human signature.
• It will not ship to mainnet before testnet. The same plan has to pass on bsc-testnet first, every time.
• It will not invent a token. Issuance is a business decision, and an agent quietly adding one to a plan is a failure mode, not a feature.
• It will not build custody.Anything that holds other people's funds at scale needs an audit and a lawyer, and neither is a thing this can do.

Things it is measurably bad at

• Novel cryptography. It will use well-understood primitives correctly and will refuse to design new ones. If your plan needs a new scheme, it will say so and stop.
• Gas golfing. Contracts come out readable and correct, not minimal. For most builds that trade is right; if you are deploying something where a few thousand gas per call matters, expect to hand-optimise.
• Design taste at the edges. It produces a clean, consistent interface. It does not produce a distinctive one, and it will not pretend otherwise.
• Judging whether people want this. It can tell you the plan is incoherent. It cannot tell you the market is.

Where a human has to stay

• Signing every mainnet transaction.
• Answering the “will not guess” questions.
• Deciding whether an overridden cut was worth it, once the thing is live.
• Anything touching real money at real scale — an audit is a separate exercise, and this is not it.

Security posture

Contracts ship with tests, and the test plan runs against testnet before a mainnet transaction is prepared. That is a floor, not an audit. Flux Ape does not claim its output is audited, and any build that would hold meaningful value should be reviewed by people whose job that is.

Back to Quickstart.