Rends decides whether an AI agent’s action proceeds — and keeps the reason

1. It decides whether an action proceeds.

Agents connected through the Rends SDK or API check each action before it runs: block, log, escalate, or pass through — based on the rules your compliance team approved. An escalated action is held until a person approves it, and that person must be someone other than the one who raised it.

2. It records why, in a form an examiner can verify.

Every decision is written to the organization’s hash chain before the answer returns, together with its reasons, the hash of the rule revisions in force, and the accountable party. Each record carries the hash of the one before it, so the whole chain can be recomputed from its first record, and an altered or missing record shows. You can see the check run at /verify/chain.

3. It knows the agent’s budget, its transaction and who is accountable for it.

• Who. The accountable party for an action is derived on the server from the verified credential: the signed-in user, otherwise the API key’s owner, otherwise whoever registered the agent. A request may restate it; it cannot choose it. Tested in production: Lab 01 — An agent can’t claim authority it wasn’t given.
• Budget. With spend budgets, an action that names a budget reserves against it in the same database transaction as its decision record; an exhausted budget blocks the action, and an escalated action holds its amount until it is approved or released.
• Transaction. With transaction authority, a decision made as a step of a business transaction takes its place in that transaction, and the organization’s transaction rules (counts, totals, separation of duties) are evaluated against what has already happened in it.