Product

The recorder, and everything you build on it

An aircraft recorder does not make the aircraft safer. It makes what happened knowable afterwards.

PromptWake does the same for AI-written code. Capture runs locally and costs you nothing to keep on; the parts above it turn that record into something you can hand to a regulator, an auditor or a court.

Book a demo
Audit Trail
An append-only record of every capture event, chained so that altering one entry breaks the chain and shows up on verification.
Discover
Tamper-evident chain
Append-only events
Signed export
Chain verification
UTC timestamps
Code Provenance
Line-level attribution that ties production code back to the prompt, the model and the tool that produced it — and keeps the unknowns honest.
Discover
Line-level blame
Commit linking
AI / human / unknown
Model attribution
Survives review
Compliance Reports
A dated report per repository or period: AI contribution, tools and models used, chain status and the methodology behind every figure.
Discover
Per repository
Per period
Report hash
Methodology section
JSON, HTML, PDF
Policy Engine
Rules in the repository: which tools are allowed, which directories need human sign-off, how long records are kept.
Discover
Tool allowlist
Protected directories
Human approval
Retention rules
Pre-commit check
Local Capture
The recorder itself. One command, a local daemon, SQLite on your own disk — no API keys and no proxy in front of your model.
Discover
One command
Local SQLite
No API keys
Secrets never stored
Works without git
Integrations
Twenty-four coding tools recognised automatically from the transcripts they already write. Switch tools and the record stays continuous.
Discover
Claude Code
Cursor
GitHub Copilot
OpenAI Codex
Windsurf, Zed and more

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

What the recorder answers

Which agent touched this file

Every captured event names the tool, the model where the transcript exposes it, the project and the file path.

When, in UTC

Events carry a timestamp and the machine they came from, so a sequence can be reconstructed across a team.

Why the change was made

The prompt that drove the edit is kept with the diff it produced, in full, rather than summarised away.

What this is not

Not DLP, CASB or SASE

We do not sit in the network path and cannot block or redact a prompt in flight. Blocking sensitive data from reaching public AI tools is Netskope, Zscaler or Microsoft Purview territory.

Not shadow-AI discovery

We see the coding tools that write transcripts on machines running the recorder. Finding unsanctioned AI use across a whole workforce's browsers and endpoints is a different product category.

Not an AI application firewall

Securing AI features your company ships — runtime guardrails, red teaming, AI-BOM — is Prisma AIRS, F5 or Lasso. PromptWake records how software was built, not how it behaves in production.

Not LLM observability

If you need traces, prompt versioning and evals for your own AI product, use Langfuse, LangSmith or Braintrust. Our subject is the codebase, not the model you are shipping.

Not an LLM gateway

We do not proxy requests, so we cannot cap a budget or stop a call at a limit. LiteLLM is open source and does that well. We report what was spent on work already captured.

Complementary, not a replacement

Those tools stop at the moment a prompt is sent: allowed, blocked, redacted, logged. PromptWake starts there and follows the prompt through to the code it produced. If your problem is upstream of that moment, we would rather say so now than in month three of an evaluation.

On AI accountability

Bring your auditors an answer instead of an estimate.

Fifteen minutes is enough to see what the recorder captures on a real repository — including how much it honestly cannot attribute.
Book a demoSee pricing