workscribe

Privacy Policy

Last updated: 27 October 2026

Overview

workscribe is a command-line tool that captures your terminal activity, groups it into work sessions, and builds summaries from it on your machine. This policy explains what data workscribe collects, how it is stored, what (if anything) leaves your machine, and your rights as a user.

workscribe is designed with a local-first, privacy-first approach. The tool has no user accounts and no registration, and since version 2.0 it makes no AI requests of its own. The only outbound communication from workscribe is a small anonymous install ping, on by default (see below). Your activity leaves your machine only if you ask your coding agent to summarize it with the Workscribe skill.

Data collected by the CLI

When the workscribe shell hook is active, it intercepts commands you run in your terminal. Before any data is written to disk, the following processing takes place:

  • Credential redaction. The raw command is scanned for known secret patterns — API keys, bearer tokens, JWT tokens, passwords passed as flags, private key material, and authorization headers. Any matched value is replaced with [REDACTED] before the event touches the database. You can add custom patterns via workscribe config set redact.extra.
  • Ignore list filtering. Commands on your ignore list (default: cd, ls, pwd, and others) are discarded silently before being written.
  • Event categorisation. Each command is assigned a category (e.g. code_commit, test_run, dependency_install) based on the base command. The raw arguments are not stored in the category field.

What is stored locally for each command: the redacted command string, its category and any structured details (such as a commit message or package names), a timestamp, the working directory, the repo and branch, the exit code and duration, and where it was captured (your shell, an IDE extension, or a coding agent's hook — IDE events also record the saved file's path). The database also holds any notes you add with workscribe note (stored as you wrote them), security alerts, and any summaries your coding agent saves with the Workscribe skill.

Where data is stored

All data you capture is stored exclusively on your local machine in ~/.workscribe/. workscribe has no cloud sync of your captured data. The one exception is a small anonymous install ping — see Anonymous install telemetry below.

LocationContents
~/.workscribe/workscribe.dbSQLite database — events, sessions, saved summaries
~/.workscribe/config.jsonConfiguration — workscribe 2.0 stores no API keys. A 1.x config may still hold one until you accept the clean-up prompt in workscribe init.
~/.workscribe/exports/Markdown files you generate with workscribe export
~/.workscribe/weekly-cache/Left over from 1.x; no longer used

The database and config file are created with 0600 permissions — readable and writable only by the owning user.

Anonymous install telemetry

workscribe sends a small anonymous ping on startup, at most once every 24 hours per install. This is on by default.

Sent:

  • A randomly generated install ID — not tied to your identity, your repos, or any account
  • CLI version
  • OS platform
  • Node.js version

Never sent: commands, events, sessions, or summaries — nothing from the sections above or below. Telemetry only runs when you invoke a workscribe command yourself — never from the background hook that silently captures each terminal command you type.

To opt out:

$ workscribe config set telemetry.enabled false

When your activity leaves your machine

workscribe summary, workscribe week and workscribe export never send your activity anywhere — summaries are built by an offline template engine on your machine.

Your activity leaves your machine only when you ask your coding agent to summarize it with the Workscribe skill. The agent reads workscribe sessions --json and sends it to its own model provider, as it does with the rest of your conversation. That data contains:

  • Event categories and structured fields — which can echo values you typed, such as ssh targets, packages run with npx, or Docker image names (redacted like everything else)
  • Commit subjects
  • Notes you added with workscribe note — free text, sent as you wrote it and not redacted, so keep secrets out of notes
  • For unclassified commands only, a short hint — the first 120 characters of the command, after redaction
  • Your last Aider prompt from a session, while capture.includeAiderPrompt is on
  • Each session's repo path and the directory most of its commands ran in

The agent may also read files in your repos for context. The data never contains full command strings, per-event working directories, IDE file paths, or credentials from captured commands (those are redacted before anything is stored).

The summary your agent writes is saved locally in the workscribe database, after the same secret redaction as captured commands, and the skill's own workscribe commands are kept out of your captured activity. Full details: workscribe.mintlify.site/privacy.

Coding agents

When you use the Workscribe skill, the data above is handled by your coding agent and its model provider under their own privacy policy and terms of service. workscribe has no control over how they handle it once it is received.

You are responsible for reviewing those terms, particularly if you use workscribe in a professional or regulated environment. If your summaries must never leave your machine, don't use the skill — workscribe summary runs entirely offline.

Website analytics

This website (workscribe.co) uses Vercel Analytics to collect anonymous, aggregated page view data. No cookies are set. No personal information is collected. The CLI tool sends a separate, small anonymous install ping — see Anonymous install telemetry above — distinct from this website's own analytics.

No accounts or personal data

workscribe does not require an account, email address, or any form of registration. We do not collect names, email addresses, IP addresses, or any personally identifiable information through the CLI tool. The only personal data that could exist locally is what you choose to type in your terminal.

Deleting your data

You have complete control over your local data. To delete everything workscribe has stored:

$ rm -rf ~/.workscribe

This removes all events, sessions, summaries, exports, and configuration. The shell hook remains in your ~/.zshrc or ~/.bashrc but will write nothing. To remove it entirely, delete the block between # workscribe:hook:start and # workscribe:hook:end in your shell rc file.

No backend for your data

There is no cloud database that stores your commands, sessions, or summaries, and no remote API that could receive them. All processing of your actual work happens locally on your device — the only time your work leaves it is when you ask your coding agent to summarize it with the Workscribe skill. Separately, workscribe operates a minimal backend that receives only the anonymous install ping described above — install counts and version/platform stats, nothing about your work.

Changes to this policy

If this policy changes materially, the "Last updated" date at the top of this page will be updated. We will not reduce your privacy protections without clear notice.

Contact

Questions about this privacy policy or how workscribe handles your data can be directed to privacy@workscribe.co.