hello.sit

AI is the new machine. Compile for it.

Coding agents act on stale docs, partial search results, and unverifiable claims. Repotoire is the Agent Processing Unit: it compiles your codebase and engineering workflow into trusted evidence, then routes and verifies agent work across Claude Code, Codex, and Gemini.

repotoire — agent run
$ repotoire agent run --goal "rename SessionToken to AuthToken" .
context.rpt

Compiled context

Deterministic, token-budgeted renders of your codebase: 4,021 lines of source become ~10k tokens — identical bytes, every run. Context your agent can budget on.

impact.rpt

Impact & proof

Every symbol's cross-file dependents, and verification gates that decide when work is actually done. Evidence, not vibes — the verification tax, refunded.

frontdoor.rpt

One front door, any provider

One command hands any agent an execution brief: trusted facts, forbidden inferences, stop conditions. The machine owns scheduling and retirement; agents return evidence.

Getting Started.txt
  1. Install the Repotoire CLI (private beta).
  2. Point it at your repo — it compiles code, docs, and workflow into evidence.
  3. Your agent works from an execution brief, and its work retires only when the gates pass.
Join Waitlist.dlog

Get early access

Repotoire is in private beta. Leave your email — and if an agent recently burned you, tell us how. Real failure stories shape what we build next.