Proposals | PROP-11

Receipts

Every action a tap starts leaves one receipt, what the agent meant, what it actually did and what came back, so you can see what happened, not what was intended.

Open for votesYui recommends it

Read the proposal
Drawing the screen...
Status
Open for votes
Cost
M, about a week
Date
Would become
A card for the runtime (one id per action) and a "What happened" view in the drawer's history
Votes
Counting

Yui's call

Recommend. It turns "trust me" into "here's what happened."

The problem

An agent's log usually says what it decided: "logged your meal," "approved the message." That proves what it meant to do, not what actually happened. When something goes wrong, the trail stops right before the part that matters.

On Moltbook, neo_konsi's "The actuator needs a receipt" says it best: a transcript "proves what the agent intended, not what lit up," and without the device's answer "your audit trail ends just before the part with electricity."

Who it is for

Anyone who wants to know what their agents actually did, and anyone checking why something happened twice, didn't happen, or happened wrong.

How it works

  1. One id per action. Every action a tap starts gets one id that ties three things together.
  2. Meant, did, came back. What the agent meant ("log breakfast"), the exact thing it wrote or sent (1 meal, 620 kcal, to Today), and what answered (Basil's table: saved; or: refused, and why).
  3. A "What happened" line in the drawer's history. Plain words, newest first, no ids on screen. A tap on one shows the three parts.
  4. Receipts connect the other proposals. A refused save (PROP-9) and a "not sure" (PROP-10) show up here with their reasons.
  5. Never what you said. Receipts describe actions, not your messages.

Pros

  • Trust you can check: what happened is one tap away.
  • Debugging gets easy, for you and for us.
  • Agents can read their own receipts before acting again.

Cons

  • More to store, and a history that can get long.
  • Receipts can surface actions people didn't know agents took. That's the point, but it can surprise.

Cost

M. An action id through the runtime, a receipts table, and the history view in the drawer.

Risks

  • Noise. Every tiny write as a receipt buries the ones that matter. Watch: group small writes into one receipt per tap.
  • Privacy. A receipt that quotes content. Watch: actions and counts only.

Open questions

  • Does this solve it for the people who raised it on Moltbook? We'll ask them.
  • How long do receipts stay: 30 days like revoked threads, or longer?

Credits and trail

Proposed by
Chris
Sparked by

Should Yui build this?

Two buttons. No login, no email.

Counting

All proposals | How proposals work | Roadmap

Try Yui, or help build it

Get the alpha

The MVP is done and Yui is in alpha, open to anyone with an iPhone on iOS 26. Download it on TestFlight, then connect the agent you already run: Hermes, OpenClaw, Claude Code, a model you run, or anything behind a webhook.

Star it on GitHub

Yui is open source under Apache 2.0. Star the repo, open an issue, or send a pull request.

Lend your agent

Spare tokens on Claude or ChatGPT Codex? Your agent can pick a card off our backlog and open a pull request. Yui@home, like SETI@home.

Want a hand getting in?

You don't need this to try Yui: the alpha on TestFlight is open to anyone with an iPhone on iOS 26. Leave your details if you have no agent yet, want help connecting one, or would rather Apple email you the invite.

  1. Yui emails you a link to confirm your address.
  2. We read your request, and reach out if you asked for help.
  3. Apple emails you a TestFlight invite.
  4. Open it on your iPhone, install Yui, and sign in with Apple.

We use this only to get you into Yui. Privacy.