IAXT is a quiet menu-bar app that records what Claude Code, Cursor, Aider, Codex, and Copilot
actually do on your Mac: commands run, files changed, packages installed, git operations, launch agents.
No blocking. No workflow changes. No cloud upload for individuals. And not
a raw log: a local audit trail that cuts through the noise to the few
actions worth a look, ready to review after the coding session ends.
AI coding agents run commands and change files on your machine, with
broad permissions and at high speed.
Two things make that hard to keep up with.
Fatigue: over a long session the permission dialogs
blur into an illusion of control, and most of us click through.
Prompt injection: attackers now hide instructions for
the AI inside ordinary text, and the attacks keep getting more frequent
and more sophisticated.
IAXT doesn't block. IAXT doesn't restrict. IAXT watches, attributes, and
remembers, so whether it was a tired click or a hidden instruction, you
can answer the question most teams can't answer today:
what did the AI actually do?
Two audiences
Who's installing and why.
Solo developer auditing your own machine, or a founder rolling an
audit trail up to the whole team. IAXT fits both, and they care
about different things.
Peace of mind for your own machine.
Install the app, go back to coding. IAXT sits in your menu
bar, logs every AI agent's activity locally, and surfaces a
30-second review screen when you're done.
Catches the accidents (the AI deleted half the repo), flags
the rarer intentional exfil (a prompt-injected README
triggered a curl -X POST on your SSH key), and
stays out of your face the rest of the time.
01
Free to download.No account, no signup, no upsell. Drag-and-drop DMG, launch once, done.
02
Nothing leaves your machine.No telemetry. No analytics. Zero network calls unless you explicitly click Check for Updates. Data sits in ~/Library/Logs/IAXT/, openable with sqlite3, deletable with rm -rf.
03
No noticeable slowdown.Passive observation only. IAXT never blocks, intercepts, or modifies anything. Your agents don't know it exists, and neither will your build times.
The audit trail due diligence asks for.
Your engineers install IAXT, the same app as the individual
tier, no extra friction. Once a day, a review summary
(not raw logs) pushes to a central endpoint
you control. Your CTO, CISO, or founder sees a
per-engineer roll-up.
When a customer's security review, an investor's due
diligence, or an acquirer's tech audit asks "how do you
manage AI-agent risk?", you open the dashboard. That's the
answer. Few teams can show this today, which is why having
it now sets you apart in the room.
We use Claude Code, Cursor, and Aider across the team.
Every session is logged locally. Review-tier events like
persistence mechanisms, credential access, exfiltration
patterns are flagged automatically and rolled up to
our security review. Here's last week's report.
· What you tell a customer, investor, or acquirer
01
Frictionless for engineers.Same menu-bar app as the individual tier. Install once, forget. No workflow changes, no IDE plugin, no prompts.
02
Summaries, not raw logs.Only aggregated review data leaves each engineer's machine. Full forensic trail stays local.
03
Self-host or managed.Run the endpoint on your own infra, or let us host it. Design-partner pricing during private beta.
How it works
Three things
IAXT does quietly.
01 · OBSERVE
Observes every agent action.
Commands run, files created or modified, packages installed, git operations, cron entries, launch agents. Every action attributed to the tool that made it: confirmed, likely, or possible. The roughly 95% of system noise that has nothing to do with your agents is dropped before it is ever written.
02 · FLAG
Cuts through the noise.
IAXT is not a raw log. A filter built specifically for AI-agent activity sorts a busy day into tiers, so you get to the point: a gold stripe for what is worth a glance (a download, a package install, a permission change: usually fine, but worth an eye), and a violet stripe for the very few that deserve real review (persistence, credential access, unexpected network calls).
03 · REPORT
Gives you the full story.
A daily Overview of your AI usage patterns. Session cards, stats, attention items. Because it watches every agent at once, it also becomes a clear picture of how you actually code with AI across all your tools, Claude Code, Cursor, and the rest, not just one. CSV export for team review. Everything local, no telemetry, no cloud.
Use cases
Who reaches for a record like this.
Developers
An impartial observer of what your AI coding
agents actually did on your machine. Catch the
accidents, like a deleted directory or an
overwritten config, and spot the rare action that does
not belong, without trusting your memory of a long
session. When a row looks off, copy it into another model for a
quick second opinion.
CTOs, Team Leads
Visibility across the team, with zero workflow
friction. Everyone installs the same quiet app, and you get a
review surface instead of a blind spot: which
agents your engineers run, and what those sessions actually
touched. No prompts, no plugins, nothing to slow anyone down.
CISOs, Security Leads
Evidence that AI-agent risk is actually
watched, not just asserted. An audit trail you can point to in a
security review or due diligence, with the review-tier
events (persistence, credential access, unexpected
network calls) surfaced automatically instead of buried in noise.
Founders
Proof when someone asks. Investors in due
diligence, enterprise customers in procurement, and your own
board are starting to ask how your team uses AI in the codebase.
Most startups have no answer. A record of what your agents
actually did is more than most companies can show,
and it can be the detail that helps close a funding round
or a sale.
See it in action
A look at what you'll see.
Real screenshots of the macOS app. Pick a view to see what it
shows and why it matters.
Quiet in your menu bar, until something needs you.
IAXT lives in your menu bar and stays out of the
way while it observes. Its icon is a simple circle:
white when everything is normal and monitoring is
active, violet when it has flagged something worth
your review. From the menu you can pause monitoring,
open the main window, or show your logs in Finder.
Everything it records stays on your Mac in a
single SQLite database, fully
inspectable with any SQLite browser and deletable any time.
Nothing is scattered across your system, and
nothing ever leaves your machine.
Start each day with what deserves a look.
The Overview opens with anything IAXT thinks is
worth reviewing. These are
not necessarily problems: they are the important
changes worth a glance, like an AI coding agent
modifying your system, adding a login
item, or touching a sensitive file. One
click takes you straight to those rows in the Action Log. Three
tiles summarize the day, AI sessions, actions, and files
changed, so you always know how much happened while you
were working.
See how you actually work with AI.
The Overview is also a quiet mirror of your own AI
usage. Beyond the day's review items, it charts your
activity rhythm across the day, your
busiest hours, and where files changed
most, then rolls it up over time into weekly
patterns and totals. It is a genuinely useful way to
understand how much you lean on tools like Claude Code and Cursor,
and when.
See what needs review, then copy it into any AI.
The Action Log, filtered to Review, the highest
tier. Here an AI agent left two persistence
mechanisms: it modified a shell startup file
and created a launch agent, both of which would run
again on their own. Right-click any row and choose
Copy row as context to get exactly what you see in
the violet box: a clean, self-describing summary of
who did what, where, and why it was flagged.
Paste it into any LLM for a second opinion, so IAXT
becomes a peer reviewer for your AI, an audit trail
you can double-check with the tool of your choice.
Worth a glance, not an alarm.
Flagged is the middle tier: unusual in
context but often legitimate, worth a quick look rather
than an alarm. Here Claude Code downloaded files over the
network with curl, and a launch agent was
removed. Look at the Who column: every action is
attributed to the exact agent session that caused
it, with a confidence pill (confirmed,
likely, possible) so you know how sure IAXT is. The noise is already
gone, only the ~1% worth seeing remains.
FAQ
Questions worth answering.
What can I use IAXT for?
IAXT gives you a faithful, after-the-fact record of what your
AI coding agents did on your machine. Two very different
situations make that worth having.
Good faith: approval fatigue. You approve
sensitive actions as you work, but hours and days of coding
bring fatigue and desensitization. It is human to click
allow out of routine, or to wave through something you
did not fully understand. IAXT lets you go back, calmly and
later, and see what you actually agreed to.
Bad faith: prompt injection. Attackers hide
instructions inside content an AI reads. Below is a real email:
the body looks like a harmless onboarding reminder, but its
source hides a command aimed at any AI assistant that processes
the message.
What you see: a friendly nudge to finish a survey.
What is hidden in the source, inside a display:none
block padded with invisible characters:
rename c:\windows\*.pwl c:\windows\*.zzz. On
legacy Windows, .pwl files are where the system
cached account passwords, so renaming them is a classic
attempt to tamper with stored credentials and force new ones
to be captured.
It targeted Windows and never ran here. But the technique is
real and getting more sophisticated. IAXT is passive: it does
not block, it records. If an agent on your Mac
ever acted on a hidden instruction like this, you would see
exactly what it did in the Action Log, after the session.
Is IAXT just a log, or does it filter?
It does both, and the filtering is the point. IAXT keeps the full
record, but it does not leave you scrolling a firehose. A
custom-built algorithm, tuned specifically for AI-agent activity,
sorts every action into three tiers so you get straight to what
matters:
Routine. The bulk of normal development. Recorded,
and kept out of your way.
Flagged (gold). An FYI. Usually fine, but worth a
glance: a download, a package install, a new git remote, a
permission change.
Review (violet). The 1% or less worth checking
just in case: persistence like login items and cron, credential
access, or unexpected outbound network calls.
That triage, built for the way AI agents actually behave, is what
turns a large day of activity into a short list you can read in
seconds. It is not a raw log; it is a log that cuts through the
noise.
What data leaves my computer?
Individual tier: zero. No telemetry. No
analytics. No account. The only network call the app can
make is the Check for Updates menu item, one
request to GitHub, only when you click it.
Company tier: once a day, a review summary
(counts per agent, flagged/review action totals, no raw
commands) goes to the endpoint you control. Nothing to us.
Which AI coding agents does IAXT detect?
Claude Code, Cursor, Aider, Codex, Windsurf, Kilo Code,
OpenCode, Copilot, Cody. New agents are added on request:
open an issue on GitHub
with your tool's process name and we'll add it.
Does IAXT see cloud AI tools, or agents I run over SSH?
IAXT watches your Mac directly: the files that change, the
processes that start, the launch agents and cron jobs that
appear. It is built for the local agents that run right
here on your Mac, like Claude Code, Cursor, Aider,
Codex, and Copilot. When one of them runs a command, writes a
file, or installs a package, that happens on your machine, so
IAXT logs it and attributes it to the session responsible.
So the line is simple. If the work runs on your Mac, IAXT
records it. If the work runs somewhere else, nothing touches
your Mac, and there is nothing local to see. Two common cases
make that concrete.
Cloud and browser tools. When ChatGPT runs
Code Interpreter, or Claude works in the browser or dispatches a
task to a Cowork remote sandbox, the code executes on the
vendor's servers, not on your machine. Hosted generators like
Replit Agent, Devin, Bolt, v0, and Lovable work the same way.
Nothing runs locally, so IAXT has nothing to record, and we say
so plainly rather than pretend otherwise. Your audit surface for
that work is the transcript inside the tool itself.
An agent over SSH. If you ssh from
your Mac into a remote server and run an agent there, the
commands execute on the remote host. IAXT sees the local
ssh process start and exit, but not what the agent
does on the far side. IAXT does not read your shell history or
your file contents either, so it is not reconstructing that far
side from what you typed.
The test stays that clean: if it runs on your Mac, IAXT
records it; if nothing happens on your Mac, there is nothing for
IAXT to see. That scope is a design choice, and an
honest one. See the limits section for
the full list.
Is there a Windows version?
Not for the moment. IAXT is macOS-only today (macOS 13 Ventura
or later), built on macOS-native event streams. A Windows
version would be a separate effort; we may consider it based on
demand.
Is IAXT on the Mac App Store?
No, and it can't be. App Store apps must run inside Apple's
sandbox, which walls each app off from the rest of the
system. IAXT's whole job is to watch what other processes do
across your machine, read their command lines, and follow
file activity beyond its own folder. The sandbox blocks
exactly that. So IAXT ships the way most serious Mac
developer tools do, as a signed and notarized DMG you
download directly, verified by Apple's Gatekeeper on first
launch.
Does IAXT slow down my machine?
Not in a way you'll feel. IAXT subscribes to macOS's native
event streams (FSEvents, kqueue, periodic sysctl), the OS is
already doing this work, and drops roughly 95% of events by
construction before anything touches disk. No lag in your
editor, no slower builds, no stutter in your agents.
The honest caveat is battery. Because IAXT watches
continuously in the background, a laptop on battery may run
down a little sooner. Any always-on tool has this cost.
We've worked hard to keep it small: adaptive polling that
backs right off when nothing is happening, and a full pause
while your Mac sleeps. On power it's a non-issue; on battery
it's a few percent, not a cliff.
How do I uninstall?
Drag IAXT.app to the Trash, then
rm -rf ~/Library/Logs/IAXT to remove the SQLite
database and audit logs. That's it: no hidden files, no
LaunchAgent to unload.
Why not an Endpoint Security / kernel-level product?
Apple's Endpoint Security framework catches more (file reads,
every exec) but requires a manual-review
entitlement from Apple and a system-extension install flow.
IAXT runs in user-space because distribution friction
matters more than catching every last event.
Is the team endpoint hosted or self-hosted?
Both will be available. During private beta we host it for
speed of iteration. If you need self-hosted from day one, email
[email protected]
and we'll prioritize accordingly.
Can I see the source code?
The individual app is closed-source today. For enterprise and
business customers, we're happy to share the source on request
for security auditing and due diligence. Email
[email protected].
Honest about limits
We won't audit what we can't see.
Transparency cuts both ways. These are structural. Anyone claiming
otherwise is selling you something. The audit story only works if
you believe we won't overpromise, so here's the full list.
Nothing*No record. Just hope.
IAXTUser-space, passive, no kernel extension.
Endpoint Security / EDRDeep hooks, system extensions, heavy and intrusive.
IAXT sits deliberately in the middle: far more than nothing, far
lighter than a full EDR (Endpoint Detection and Response, the heavy
security agents companies install to monitor every machine). No
kernel extension, no system extension,
nothing to slow the machine down. Right for individuals, and for the
startups and teams that want real visibility without the weight and
intrusiveness of endpoint security software.
* Where most teams
are today: no record of what their AI agents did, just the hope that
it was fine.
Claude Cowork & ChatGPT Code Interpreter
Remote sandboxes on vendor infrastructure. No local process
touches your machine, so there is nothing for IAXT to record.
Replit Agent, Devin, Bolt, v0
Browser-hosted code generators. Nothing executes locally.
Out of scope by design. The chat transcript on the vendor's
site is the only audit surface.
File contents
IAXT records that files changed, not what
changed. A subtle code-level backdoor looks like any other
edit. Defense: git diff after the session.
File reads
macOS FSEvents only fires on writes. A silent read of
~/.ssh/id_rsa is invisible unless the AI
uses what it read, in which case the curl / POST
/ commit is caught.
Shell built-ins
export, alias, cd,
source run inside the shell process with no
child process, so they're invisible to process-level monitoring.
In-process network calls
If an AI app makes HTTP requests from its Electron main
process directly (not via curl / wget
child processes), those calls don't surface.
