Block before it publishes, not after.

A synchronous allow or block decision, called before content goes live. Checking after the fact tells you what already went wrong. This tells you before it does.

What this actually is

A decision API, called by your code

Your system sends content to /api/v1/enforce at the moment it would otherwise go live, publishing a post, letting an agent proceed, sending a message. It gets back an allow or block decision, a risk score, and the specific flags behind it, in milliseconds, no external API calls in the path to slow it down. Your own code makes the final call on what to do with a block: hold the content, escalate to a human, or reject it outright.

What this is not

Not a network proxy

This is not a transparent gateway that sits inside your infrastructure intercepting every request automatically. That is a much bigger, much riskier piece of engineering, one that has to guarantee near constant uptime and sub second latency for every single call your production systems make, because a fault on our side would then be a fault in your live path. We are not claiming that, and anyone who tells you a small team ships that lightly is worth asking hard questions of.

What we do claim: a fast, synchronous decision your own code calls voluntarily, at the point you choose to call it. The difference is not just words, it is where the operational risk actually sits, with you, deciding when and how to call a gate, rather than with a middleman positioned to break your production traffic if it stumbles.

What gets sealed

Blocks get evidence, not just a log line

Every blocked decision is sealed with an independent RFC 3161 timestamp and a public verify link, the same mechanism behind every other sealed record on this site. That means a block cannot quietly be edited or deleted after the fact to look like it never happened, or never existed to look like it did. Allowed decisions are recorded too, for your own review, but are not individually timestamped, to keep routine traffic from flooding the chain.

Wire it in

Full request and response shape in the API docs. Needs a Red Flag AI Pro API key, generated free from Settings.