Getting started
Tracemason records what AI agents do — sessions, tool calls, file changes, commands — evaluates that activity against security policies, and links it to the Git history it produced.
This guide takes you from nothing to a real Claude Code session on screen. There is nothing to host and nothing to clone: you need an account and one command.
Tracemason is pre-release software. This documentation describes what is implemented today; anything not yet built is marked planned.
1. Create your organization
Go to app.tracemason.com/sign-up, create an account and name your organization. You become its Owner. The dashboard starts empty.
2. Connect Claude Code
On the machine where you use Claude Code, run:
npx @tracemason/cli init
Pre-release:
@tracemason/cliis being published to npm. Until it is listed there, this command will not resolve.
This needs Node.js 20 or later. The command opens your browser: sign in, pick the organization, approve. It then installs the Claude Code hooks, sends a test event and prints the dashboard link.
Nothing else has to be installed. The collector copies itself to ~/.tracemason/bin, so it keeps working after the npx cache is cleared. See Claude Code collector for exactly what it does and What is collected for what it sends.
3. Watch a session
Start Claude Code in any Git repository and ask it to do something. Open Live Activity — events arrive as they happen. Open the session under Traces to see the timeline.
Adding teammates
Invite them from Settings → Members. Each developer then runs the same command on their own machine and approves it for the organization:
npx @tracemason/cli init
Removing it
npx @tracemason/cli uninstall --purge
This removes the hooks from your Claude Code settings and deletes ~/.tracemason.
Next
- Concepts — sessions, events, findings, provenance
- What is collected — and what never is
- Security policies
- GitHub integration
- Self-hosting — run Tracemason on your own infrastructure