CAIN Swarm Guard, explained for beginners

Swarm Guard puts one line of code between an AI agent and each tool, so a risky call stops before the tool runs.

What it is

AI agents call tools: read a file, send a payment, delete a record. Usually nothing stands between the agent's decision and the tool.

Swarm Guard wraps each tool in one line of code. An unknown tool, a destructive argument or a declared high risk does not run; it waits for a person. In a group of agents, a handoff that tries to widen authority or extend its lifetime is refused.

An everyday comparison

A guard at each door who checks the list and looks in the bag, and a rule that nobody can lend out a key they don't hold.

How it works, in 4 steps

1. The tool is wrapped with @cain.guard.
2. The agent calls the tool.
3. The guard checks the allowlist, the arguments and the risk.
4. Allowed: the tool runs. Otherwise: the tool body is never reached.

Try it now

Open the eight live demo tools → Every button sends a real request to the live product and shows the real answer.

See it on the homepage: CAIN Swarm Guard section →

Who it is for

Teams building agents in Python that call real tools.

FIRME AI CORPORATION · CAIN-42 is a Byzantine trust fabric. Numbers marked "our own check" were run by us, not by an independent party. Nothing on these pages is a certification.