Skip to main content

​
Privacy & Data Safety

auto-sop is built on three privacy principles. Understanding them takes 60 seconds.

​
The three principles

Captures never leave your machine

Raw tool-call captures, session transcripts, and your source code stay local. Always. No exceptions.

Your own Claude does the learning

LLM analysis runs through your existing Claude Max subscription. auto-sop adds $0 in LLM costs — you’re already paying for Claude.

Cloud only gets previews + stats

The server receives the first 10 words of each directive and aggregate statistics. Nothing else.

​
What stays local

Everything sensitive lives on your machine and never touches a server:
auto-sop never reads your source code. It only captures Claude Code tool calls and their outcomes — the same data Claude Code already processes locally.

​
What syncs to the cloud

Cloud sync is available on all plans and is always opt-in via auto-sop install --license <key>. The Free plan syncs stats and project data for 1 project. The Pro plan adds directive sync and unlimited projects. When cloud sync is active, the server receives:
Directive previews are truncated to exactly 10 words before leaving your machine. The full directive text is never transmitted.

​
Free tier — minimal cloud footprint

On the free plan with a license key, auto-sop communicates only:
  • License validation — periodic key check (encrypted)
  • Stats sync — aggregate metrics (tokens saved, errors prevented) for 1 project
  • Project binding — project slug and machine ID
No directive content is synced on the free plan (directive sync is Pro only). Without a license key, auto-sop operates fully offline with zero network requests.

​
Source-available — Elastic License 2.0

auto-sop is source-available under the Elastic License 2.0 (ELv2). You can audit the source code to verify these privacy claims:
  • The hook shim that captures tool calls
  • The learner that analyzes patterns
  • The sync module that sends previews to the cloud
  • The managed-section writer that updates CLAUDE.md
Don’t trust, verify. The entire capture and sync pipeline is in the open-source repository. Search for syncStats to see exactly what leaves your machine.

​
Data flow diagram

​
Questions?

See the FAQ for common privacy questions, or open an issue on GitHub.