GitScribe

Privacy Policy

Last updated 16 September 2026 · applies to the GitScribe Chrome extension and the GitScribe MCP server

There is no GitScribe server. We operate no backend, no account system and no analytics, so there is nothing of ours in the path to collect anything with. Your settings stay in your browser profile. One feature — generating a README — sends the code you selected to a model provider you choose, and it is the only one that sends your code anywhere.

What we collect

Nothing. GitScribe has no telemetry, no crash reporting, no usage metrics, no advertising identifiers and no cookies. We do not see which repositories you open, what you select, or that you installed it at all. We do not sell or share data, because we never receive any.

What is stored, and where

Everything GitScribe saves lives in your own browser profile, through Chrome's storage API. It is readable only by this extension, is removed when you uninstall it, and is never uploaded.

StoredWhy
Forge access tokens (GitHub, GitLab)Optional. Raises rate limits and reaches private repositories. Sent only to the forge the token belongs to, and never attached to any other URL.
Model provider API keysOptional. Used for exact token counts and README generation. Sent only to that vendor.
Selections, exclusion presets, recent repositoriesSo a selection survives navigating around a repo. History is capped at 20 entries.
Model choice and themePreferences.

Digests and generated READMEs are held in memory in the background worker and are lost when it restarts. Nothing is written to disk unless you press Save as file.

Where data goes on the network

GitScribe talks to these hosts, and no others:

Generating a README sends your selected source code to the model provider you picked. Everything else in GitScribe keeps your code on your machine; this feature cannot — the model has to read the code to describe it. You are shown the file count and the estimated cost, and nothing is sent until you confirm. Your code is then handled under that provider's privacy policy and data-retention terms, not ours. If that is unacceptable, run the analysis against a local Ollama model, which never leaves 127.0.0.1, or use the raw digest, which needs no model at all.

Permissions, and what each is for

GitScribe injects no content script into any page, and loads no remote code — everything that runs is inside the published package.

Secrets found in your code

Digests are scanned for credential-shaped strings, which are redacted before the output is assembled, and what was redacted is reported so the redaction is visible rather than silent. This is a safety net, not a guarantee: review a digest before pasting it somewhere public.

Children

GitScribe is a developer tool, is not directed at children, and collects no personal information from anyone.

Changes to this policy

Material changes will be published here with a new date, and — because the policy ships in the repository — visible in its git history.

Contact

Questions or a suspected privacy problem: open an issue at github.com/OWNER/gitscribe/issues.