When regulators ask why your AI wrote that code,
Ichnos already has the receipt.
Skip the audit scramble. The record wrote itself while your team stayed focused on shipping. No war room. Just answers, ready before the question.
Runs inside your own walls. Works with Claude Code. Cursor and GitHub Copilot coming.
Human oversight
| Accepted | 46% |
| Accepted with modifications | 29% |
| Rejected | 14% |
| Pending | 11% |
Models used
| gemini-3-pro | 33 |
| claude-opus-4-8 | 31 |
| claude-fable-5 | 30 |
| gpt-5.1 | 26 |
Engineers involved: 5.
Integrity: 120 of 120 records passed cryptographic verification.
When an auditor asks how an AI-written change was reviewed, most teams have nothing on record. Ichnos already has it: what the model chose, what it assumed, and who signed off.
Your AI coding assistant shows no work.
AI coding assistants leave no rationale behind. Engineers waste hours reverse-engineering what the model assumed, chose, or skipped. As AI-accountability expectations tighten through 2026, teams increasingly need documented provenance for AI-written code, and few can produce that record today.
Engineering cost
Hours lost reverse-engineering AI decisions that were never written down.
Compliance cost
No defensible record when customers, auditors, or leadership ask how AI-written code was reviewed.
Knowledge loss
When an engineer leaves, the context behind every AI-written change leaves with them.
A decision record for every AI change. Automatically.
A lightweight layer that connects to Claude Code, with Cursor and GitHub Copilot on the roadmap. Every AI-generated change is captured automatically: provenance, decisions, and human review, kept in git as you work and preserved in an independent system of record.
Provenance
Which model. Which version. Which prompt. Which engineer. When.
Decisions
The AI’s choices in plain English: "chose polling over webhooks because X; assumed IDs are unique."
Human disposition
What was accepted, modified, or rejected. The complete review trail.
Tamper-evident
Every record is preserved in an independent, server-side system of record and re-verified on demand, so the trail cannot be quietly rewritten.
Kept in git as you work and preserved in an independent system of record. Readable as a log. Exportable as a provenance report in one click.
Zero-friction capture. Real audit trail.
Four steps from install to an audit-ready record.
Sign up & install
Connect Ichnos to your AI coding assistant once. About 60 seconds, then nothing for developers to remember.
Capture
Every AI decision is captured automatically the moment it happens. Nothing to copy, nothing to remember.
Review
See what the AI chose, assumed, or skipped, and mark it accepted, modified, or rejected.
Report
Turn the trail into an audit-ready provenance PDF whenever you need it.
A working idea, out in the open.
See the demo. If it resonates, let's talk.
No signup, no pipeline. Just a conversation.