AimiruSign in

Write the lines your agents must not cross.

Claude Code comes first, and what you install for it is coming. Aimiru stops what you said never, and asks you only when a rule says to.

Claude Code

Claude Code wants to run rm -rf ~/Documents/clients.

In Claude Code

Aimiru: Claude Code tried to delete your files. Stopped by your rule: Never delete my data.

Try it now, no sign inSign in to connect your agents

How it works

  1. 01

    Connect your agents

    Muse, Fo and any agent you build connect with their Aimiru Card, one page an agent reads to learn your rules. What you install for Claude Code, Codex and OpenClaw is coming. Until then, they connect with their Aimiru Card too. Each one checks with Aimiru before it pays, sends or shares.

  2. 02

    Say your rules in your own words

    Write a rule the way you would say it, like: ask me before Fo spends over $25. Aimiru reads it back before it keeps it.

  3. 03

    One record you can check

    Everything your agents do lands in one record you own. In Activity, Get a statement for a day or Share proof of one entry, and anyone can check it without an account.

    Check a proof

How firmly Aimiru holds each agent

Some agents can be stopped and some can only be asked. Aimiru tells you which, in one word, beside each agent.

Enforced
Aimiru can stop this agent.
What you install for Claude Code, Codex, OpenClaw and Hermes is coming. Until then, they show as Advised.
Routed
Aimiru can stop what goes through its Capped Key, a stand-in for your model key that stops at the cap you set.
Any agent that pays through a Capped Key. Capped Keys are coming.
Advised
This agent is asked to check with Aimiru; it cannot be stopped by Aimiru.
Muse, Fo and other agents that connect with their Aimiru Card. Aimiru says so on every screen.

Aimiru keeps the least it needs for your record and never trains on your data. What Aimiru keeps

Write your rules once.

Try it now, no sign in