← klaro.services
SentinelConsentraKlaroShieldBundlesPricing

Privacy & Telemetry

This exists to earn your trust, not to hide what the SDK does. Read it, or just run npx klaro telemetry status yourself.

This is not analytics on you.

@klaroshield/sdk sends anonymous, aggregate product telemetry -- installations, active projects, feature adoption -- so we can understand whether KlaroShield is actually useful and where to invest next. It is not tied to your identity, your account (there isn't one), or your company. No registration, no login, ever required to use the full SDK.

What is collected

✓ Anonymous installation ID
A random UUID generated on first run, stored in .klaro/install.json. Never derived from your hardware, hostname, or MAC address.
✓ Anonymous project ID
A random UUID per project, created by klaro init. Not tied to any account -- there is no account.
✓ SDK and Node.js version
So we know which versions are actually in use when deciding what to support.
✓ Platform
darwin / linux / win32 -- nothing more specific.
✓ Which middleware is configured
That you've enabled retries()/budget()/pii()/etc -- not their options, not what they redacted.
✓ Which CLI command ran
doctor, stats, simulate, benchmark, dashboard -- so we know which tools are actually used.

What is never collected

  • ✗Prompts you send to any AI provider
  • ✗Responses from any AI provider
  • ✗API keys or any secret
  • ✗PII of any kind
  • ✗The contents of .klaro/logs.jsonl or any local file
  • ✗Your source code
  • ✗Anything from a request or response body

How to disable it

Telemetry is opt-out, not opt-in -- but it's a single command either way, and the SDK is fully functional with it off:

npx klaro telemetry disable

# or, to override .klaro/config.json entirely:
export KLARO_TELEMETRY=0

Why we collect anything at all

Because we're building this in the open, without funding a guess. Knowing that klaro doctor is used ten times more than klaro benchmark, or that most projects enable secrets() but skip validation(), tells us where to spend the next month of engineering time -- instead of guessing, or making every developer fill out a survey. If that trade-off doesn't work for you, turning it off costs you nothing: every feature of the SDK works identically either way.