Open source · MIT · works with Claude Code, Codex & more

Undo agent file changes

Run Claude Code, Codex, or any hooked agent with a local record of what changed. Stord logs file reads and writes by agent and lets you reverse changes from the CLI.

See it in action

Watch Stord log and undo changes

In this demo, one agent reads .env and another deletes build/. stord log shows the timeline, and stord undo restores the files.

stord · agent file log
Terminal demo: stord log shows a Claude Code agent reading .env and a Codex agent deleting build/, then stord undo restores the deleted files and stord agents lists every identity.

The problem

Coding agents can change files faster than teams can review them.

Teams often run agents with broad local permissions. If an agent overwrites a config, deletes a directory, or reads a secret, the default logs rarely show the file-level timeline.

Identity tools know who the agent is. Gateways know what it called. Stord records what happened to your files and keeps a restore point for writes.

63%of orgs can’t stop agent work quickly when something goes wrong
33%have no audit trail of agent activity at all
45:1machine identities now outnumber humans

What Stord does

File-level logging and rollback for coding agents.

Use the open-source CLI locally today. It records reads and snapshots writes with no hosted dependency. The hosted version adds enforced path permissions, off-host logs, and a dashboard for teams.

01 · AUDITAvailable now

File activity log

Records each read, write, and delete with agent, session, and user context. stord log shows the timeline; stord agents shows who touched each path.

02 · REVERSEAvailable now

The undo button

Stord snapshots a file before an agent writes to it. Use stord undo for one operation or stord restore for a broader rollback, including deleted directories. Git history stays untouched.

03 · PERMISSIONOn the roadmap

Scoped agent credentials

Issue keys to agents instead of teams. Allow paths such as write: /build/**, enforce them in the hosted service, and review off-host audit logs from one dashboard. Join the waitlist →

Free & open source

Install it in one line.

A single Node CLI with zero runtime dependencies and an MIT license. It works with Claude Code today and with Codex, Hermes, OpenClaw, or custom agents through hooks.

# in any project your agents work in
$ stord enable
$ claude --dangerously-skip-permissions
$ stord log  # everything it touched
$ stord undo last  # take it back

Cloud early access

Manage agent access across machines.

Use the CLI locally today. The hosted service adds per-agent permissions, off-host audit logs, and one dashboard for every agent and machine. Join the waitlist for early access.

Launch updates only.