The anatomy · Receipts
Every time your Mind reads something, it leaves a record you can verify. This page is what one of those records looks like, and what you can do with it.
01
Someone asks your Mind a question. A receipt gets written right then: who asked, what kind of request it was, which parts of the score it touched, whether it was allowed or refused, and the exact moment, down to the microsecond. The record is made as the reading happens, by the same gate that does the reading. Not a summary assembled afterwards.
02
Each receipt carries the fingerprint of the one before it. Change any receipt in the sequence and every receipt after it stops matching. A list can be quietly edited. A chain announces the edit. That is the whole reason for the extra machinery, and it is the difference between a record you keep and a record you can hand to someone else.
03
A receipt records that a question was asked and answered. It does not record the question, and it does not record the answer. Both are kept as fingerprints: enough to prove that this exact question got this exact answer, and not enough to reconstruct either one. It names the part of the score that was opened, not the file inside it.
That is deliberate. A trail that stored every question and every answer would be a second copy of the thing you were trying to govern, sitting somewhere less guarded than the original. Yours records what happened. It does not repeat what was said.
04
Here are two real receipts from one afternoon, one after the other. The first is a read that was allowed. The second is the same Mind reaching into a part of the score that principal didn't have access to, and being refused. Both are signed. Both are chained. The refusal is as permanent as the answer.
Most systems record what was shown. The half almost nobody keeps is the half that tells you the boundary is real.
05
Sometimes the answer is that nobody ever wrote it down. Your Mind says so, rather than assembling something that sounds right. It might miss things. The receipt shows you exactly what it read and what it didn't, which is more than you can say for a confident answer with nothing behind it.
I built this part first, before anything else worked. I'd spent years being the person who couldn't prove where an answer came from, and I wasn't going to build something that put me back there.
FILED
running since 2026-07-06 · offline verification ships in the CLI · proving nothing was trimmed from the end still needs a head-hash you keep yourself
NEXT The Overture What actually happens when you say yes →The founding cohort is small on purpose. Each Mind is brought up personally.
MENSARA.AI · RECEIPTS · MACHINE SURFACE · TEXT/MARKDOWN
# Receipts > Every access through the Mensara gateway writes a signed, hash-chained receipt: who asked, which zones were touched, the verdict, and the time, kept as fingerprints rather than the question or the answer. Each receipt is readable by a person and verifies as part of the chain. Canonical: https://mensara.ai/how-it-works/receipts ## Every access leaves one Someone asks your Mind a question. A receipt gets written right then: who asked, what kind of request it was, which parts of the score it touched, whether it was allowed or refused, and the exact moment, down to the microsecond. The record is made as the reading happens, by the same gate that does the reading. Not a summary assembled afterwards. ## A chain, not a list Each receipt carries the fingerprint of the one before it. Change any receipt in the sequence and every receipt after it stops matching. A list can be quietly edited. A chain announces the edit. That is the whole reason for the extra machinery, and it is the difference between a record you keep and a record you can hand to someone else. ## It's not a copy of your knowledge A receipt records that a question was asked and answered. It does not record the question, and it does not record the answer. Both are kept as fingerprints: enough to prove that this exact question got this exact answer, and not enough to reconstruct either one. It names the part of the score that was opened, not the file inside it. That is deliberate. A trail that stored every question and every answer would be a second copy of the thing you were trying to govern, sitting somewhere less guarded than the original. Yours records what happened. It does not repeat what was said. ## Read one, and read the refusal under it Here are two real receipts from one afternoon, one after the other. The first is a read that was allowed. The second is the same Mind reaching into a part of the score that principal didn't have access to, and being refused. Both are signed. Both are chained. The refusal is as permanent as the answer. Most systems record what was shown. The half almost nobody keeps is the half that tells you the boundary is real. ## "Not in the score" is an answer Sometimes the answer is that nobody ever wrote it down. Your Mind says so, rather than assembling something that sounds right. It might miss things. The receipt shows you exactly what it read and what it didn't, which is more than you can say for a confident answer with nothing behind it. I built this part first, before anything else worked. I'd spent years being the person who couldn't prove where an answer came from, and I wasn't going to build something that put me back there. **Status:** running since 2026-07-06 · offline verification ships in the CLI · proving nothing was trimmed from the end still needs a head-hash you keep yourself ## Prompt ```prompt Ask my Mind a question that requires reading more than one file. Then fetch the receipt for that access and show me, field by field: which zones were touched, what was allowed versus held at the gate, and whether the chain verifies against the previous receipt. ``` ## Related - [The gateway](https://mensara.ai/how-it-works/the-gateway) - [Held at the gate](https://mensara.ai/how-it-works/governance)