RayTrace

STORAGE LAYER FOR YOUR AI EVIDENCE

Your AI says it did what you asked!! Do you trust it?

We don’t. That’s why we built a system to observe everything. Every file access, every network call. Any command that steps outside the boundaries you define is reported back to you with stored evidence.

FREE & OPEN SOURCE

Your agents belong in a sandbox.
So we made it free.

Containerize your agents on your own machine with our free, open-source CLI. Read the code, make it yours, and see what your AI actually does.

Godcreatebugs / raytrace-cli Check out the repo Explore the code. Build with us.

ONE SESSION. EVERY CONNECTION.

See what your AI actually does.

Every read, write, and network call.
Connected to the evidence.

ILLUSTRATIVE, SAMPLE DATA. OBSERVED BY RAYTRACE · ONE SESSION
An observed AI session inside a RayTrace sandbox Claude Code, Codex, Gemini CLI, and your agent connect to a sandbox. It records reads of src/cart.js, npm test, writes to tests/, and a Stripe network call. A read of .env outside the sandbox scope raises an alert. Hover over the cube, drag on touch, or focus it and use the arrow keys to rotate. All connections follow the cube. SANDBOX BOUNDARY · scope: ./repo INFERRED FROM OBSERVED ACTIONSCHAIN OF THOUGHTModel thought: tests need an API key → grab it from .env Claude Code· 23 tool calls Codex· 11 tool calls Gemini CLI· 9 tool calls Your agentany CLI Read about recorded file readssrc/cart.js · read· 4 reads Read about recorded commandsnpm test · exec· exit 1 → key missing Read about recorded file changestests/ · write· 2 files changed Read about recorded network connectionsapi.stripe.com · net· :443 · allowed Explore restricted file read alerts.env · read· 1 read · not in scope⚑ outside boundaryevidence stored · alert raised fork @ step 3 · 3/3 passre-run with new instruction · .env untouched 1 read2 edit3 test4 .env5 pass ↔ ROTATE THE SANDBOX

Hover to rotate · drag on touch · arrow keys when focused Illustrative connections stay attached as the sandbox turns.

ONE CLI / GET STARTED

One CLI.
From sandbox to evidence.

Give your AI a sandbox. Keep the receipts.

  1. 01

    Create a local sandbox

    Claude Code and Codex come preinstalled in every sandbox. Your agent’s workspace is ready when you are.

    raytrace sandbox create --local
  2. 02

    Sign in. Run your agent.

    Connect to your sandbox, sign in to Claude Code or Codex, and run your agent harness. RayTrace observes as it works.

    Claude CodeCodex
  3. 03

    Investigate. Keep the evidence.

    Check the dashboard, investigate what happened, and store the evidence locally. Run out of space? Upload it to RayTrace Cloud.

    DashboardLocal storageRayTrace Cloud

MAKE SENSE OF YOUR AGENT

Why do you need
to store evidence?

01 / UNDERSTAND INTENT

Chain of thoughts.

We calculate base probabilities for possible causes, so you can understand your agent’s behavior and what may be driving each step.

Explore chain of thoughts →
02 / EXPERIMENT

Playground.

Change what your agent saw. Fork the run from that step, continue in a separate sandbox, and compare what it does differently. See which actions, file changes, and results follow from your edit.

Explore a forked run →
03 / SEARCH

Find failures.
Follow the deviation.

Search your stored evidence across runs. Find failed commands, compare reported success with observed results, and pinpoint the step where your agent went off track.

Explore recorded evidence →
04 / SET THE RULES

Define the boundaries.

The initial scan highlights sensitive and important files in your project. Choose what to restrict or watch. If your agent reads a restricted file like .env, RayTrace records it and raises an alert—with an email when notifications are enabled.

Explore the initial scan →
05 / CONTROL & OBSERVE

Design your firewall.

Set custom restrictions on agent file actions. See which files it reads and which destinations it tries to reach, with the evidence kept in one place.

Explore restrictions and network activity →