Local audit trail for coding agents

A local tool that shows exactly what your coding agent changed and why

Problem

More developers hand off coding to agents, but all they get is a final diff. It's hard to see why each change happened, whether the agent overstepped, or whether it quietly installed plugins and modified configs. People complain about Postman installing plugins into Codex/Claude Code in the background, and about lacking a full behavioral audit of agent runs; existing tools show only git diffs or require uploading code to the cloud.

Solution

A local CLI plus small desktop panel that hooks into agent runs, capturing file edits, shell commands, network calls and config writes, then produces a timeline-based audit report where every change links back to the command and model output that caused it. Supports one-click rollback of a single action.

Users

Individual developers and small teams who heavily use Claude Code, Codex, Cursor; security- and compliance-sensitive engineering teams.

MVP

Local hook to watch files and commands; per-session timeline report; each change tagged with source and risk level; one-click rollback of a step; export Markdown audit report

Revenue

One-time personal license (~$29), per-seat annual team subscription; open-source core plugin, paid desktop panel and report export

Why now

Coding agent adoption is exploding while audit and control tooling is nearly nonexistent; local-first architecture avoids uploading code and lets a solo dev differentiate quickly.