About

A recorder for the code you did not type

Aircraft got recorders because investigations kept ending in speculation.

Software is arriving at the same point. A growing share of production code is written by agents that keep no durable record of what they did or why, and the organisations that carry the liability cannot reconstruct it afterwards. We think that gap closes the same way it closed in aviation: by recording as a matter of course, before anybody asks.

See the product

What we believe

Liability follows the code, not the tool

When AI-written code causes harm, the vendor of the model will not be the one explaining it. The organisation that shipped it will.

A measured number beats a confident one

We would rather report that forty percent of a file is unattributable than publish a tidy figure we cannot defend.

Recording should not require trust

Capture is local by default. A tool that demands your source code in order to tell you who wrote it has the incentives backwards.

Built on something developers already keep running

PromptWake started as a memory for AI-assisted work: one command, a local daemon, and a searchable history of every prompt, response and diff. The audit layer sits on top of that, which matters — a compliance tool nobody runs produces no evidence. The free, local developer experience stays free and local.
Local capture

From the people already running it

The rollback points alone are worth it. Claude rewrote half my state layer at 2am — I restored the pre-refactor snapshot with one click.
Jonas Weber
Indie hacker
We run it self-hosted for the whole team. Compliance asked how AI-generated code enters our repo — for the first time, we just showed them.
Priya Raman
Eng Manager, healthcare SaaS
I stopped keeping a prompts.txt file the day I installed it. Three weeks later I traced a subtle bug back to one Cursor prompt in about ten seconds.
Maya Okafor
Staff Engineer, fintech
The rollback points alone are worth it. Claude rewrote half my state layer at 2am — I restored the pre-refactor snapshot with one click.
Jonas Weber
Indie hacker
We run it self-hosted for the whole team. Compliance asked how AI-generated code enters our repo — for the first time, we just showed them.
Priya Raman
Eng Manager, healthcare SaaS

How we handle what we do not know

Attribution by content matching has real limits. Code written before the recorder was installed cannot be attributed to anyone. Heavily reformatted code stops matching. A line that appears in several interactions is resolved by adjacency and recency, which is a rule rather than a proof.

Every one of those ends up in an unknown column that we report and define. It would be straightforward to fold unknowns into the human figure and produce a cleaner chart. It would also make the number worthless to the only people who need it.

Talk to the people building it.

We are more interested in the repositories where the answer is uncomfortable than the ones where it is tidy.
Book a demoSee pricing

On AI accountability