Teams and community / You and your team
What Source tracks, and how to turn it off
You use Source every day, and you want to know what it records about you. This page lists what Source records, who sees it, how long Source keeps it, and the switch that turns it off.
8 min read
Source records which features people and agents use. This usage data goes through the Source site to Source's own analytics database (ClickHouse), on one server in the EU (Germany). No outside analytics service gets it. The usage data never holds your code, your file names or the names of private repositories.
Turn it off
- Open Settings, then Profile.
- In the Privacy card, turn off Share usage analytics.
Source saves your choice on your account, so it works on all your devices. When it is off:
- The browser stops sending usage data at once, with no reload.
- Source stores none of your events, from the browser or from the server. It drops them before they reach the database.
The switch does not delete the events that Source stored before you turned it off. Source deletes each of them when it is 2 years old.
Tip
Your browser settings can stop the browser events too. If Do Not Track or Global Privacy Control is on, the browser sends no usage data. The server does not read these settings, so to stop the server events too, turn off the switch.
If you are not signed in, Do Not Track and Global Privacy Control are the way to turn it off.
What the browser sends
Unless you turned it off, the browser sends these events to /_e on the Source site:
- Page views. When Source loads in your browser, the first page view also has the site that sent you, for example
https://example.com. - Page leaves. Each one has the time on the page, the time that the page was visible, and how far down you scrolled.
- Clicks, form submits and input changes. The event has only the tag of the element, its role, its
data-trackname and a short path of tags. It never has the text, the link or the value of the element. - Web vitals. Once for each page load, the browser sends LCP (how fast the main content appears), INP (how fast the page answers a click or a key) and CLS (how much the layout moves).
Each event names its page by the route pattern, for example /@{$owner}/$repo/blob/$. On a repository page, each event also has the id of the repository. When the repository is public, the event also has the name of the repository and the path of the page, for example /@TanStack/router/blob/main/README.md. The path never has the query part of the URL. The events of a private repository never have its name or its paths.
There is no session replay and no heatmap.
What Source adds to the browser events
When the browser events arrive, Source adds:
- The country of the request, and the Cloudflare data center that took it.
- The user agent of your browser, which names the browser and the operating system.
- The time that the events arrived.
Source never stores your IP address. It uses the address only to limit how many requests one address can send.
Who you are in the data
Each browser event has an id for who sent it, and a random session id:
- Signed in: your user id. The browser tab keeps the session id in
sessionStorage, so a full reload does not change it. - Signed out: a random id that the browser makes. The browser keeps this id and the session id in memory only, with no cookie and no storage. A full reload makes new ones.
A new session starts after 30 minutes with no event, and when you sign in or out.
What the server sends
The server sends an event for each action in the table, for example repo.created when you create a repository. When you are signed in, the event has your user id. When you are not signed in, it has anonymous.
The other fields are ids, numbers, flags and fixed labels, for example the sign-in method. An event has a repository name or a skill name only when it is public.
| Area | Events |
|---|---|
| Accounts | user.signed_up, user.signed_in, user.signed_out, token.created, org.created, org.member_invited, org.member_joined |
| Repositories | repo.created, repo.imported, repo.forked, repo.deleted, repo.visibility_changed, repo.visit |
| Git | git.clone, git.fetch, git.push |
| Issues and pull requests | issue.opened, issue.closed, pr.opened, pr.closed, pr.merged, pr.review_submitted, comment.created |
| Community | community.post_created, community.answer_created, community.answer_accepted |
| Agents, CI and the factory | agent.session.started, agent.session.ended, pr_chat.turn, arch_chat.turn, ci.run.started, ci.run.finished, factory.run.finished, factory.agent_toggled |
| Skills and integrations | skill.published, skill.installed, skill.starred, github.sync_connected, github.sync_disconnected, mcp.connection_added, vault.secret_set |
REST, MCP and src | api.call for each REST call and each MCP tool call, and cli.command for each REST call from src |
What some of these events hold:
api.callhas the channel, the name of the method or the tool, and the kind of caller. The kind of caller is a person (user), an agent (agent), a CI job (job), an app (app) or a signed-out caller (anonymous). It never has an owner or a repository.cli.commandhas the version ofsrc. Source knowssrcby its user agent,src/0.0.0.vault.secret_sethas a count only, never a name or a value.repo.visitis one view of a repository page. The events of the rankings also have the flagrank, the number of the issue or pull request, and the user id of its author. See Rankings use the same data.
What is never in the usage data
- Your code, file paths and branch names.
- The names of private repositories and private skills.
- Email addresses, tokens and secrets. The server drops each field that looks like one.
- The text of buttons and links, where links go, and what you type in a field.
- Your IP address.
Who sees the usage data
The site admins of Source see the usage data on the admin page /admin/analytics. The page shows totals and top lists only:
- Totals, for example active people, active repositories, sessions, and calls from agents and from people.
- Top lists, for example the top pages, countries, features and repositories.
- The top list of repositories also shows the names of private repositories to the site admins.
The page shows no list of people and no events of one person. The people who run the Source server can also read the database on the server.
The maintainers of a repository see the reading time and the countries of their repository from the usage data, on its Insights tab. See Repository traffic is separate.
How long Source keeps the data
Source keeps each event for 2 years. Then the database deletes it automatically.
Every night, the server saves a backup in Source's own storage on Cloudflare. Source keeps each backup for 8 days. As a result, a copy of an event can stay for up to 8 days after the database deletes it.
Rankings use the same data
The repository rankings and the monthly awards count your visits, reading time, clones, forks and contributions from this usage data. The rankings and the awards of repositories show counts only, never who you are.
Your achievements and the Top contributor medals come from the same data, and they are about you:
- Public: your achievements and your medals show on your profile, and your ring shows next to your name in threads. A medal can also show on its repository and on Explore. There is no switch to hide them.
- Opt-out: when you turn off Share usage analytics, your actions stop counting in the rankings, the awards, the medals and the achievements. The steps that you have stay.
Repository traffic is separate
The Insights tab of a repository shows its maintainers the clones and the page views. This count is not part of the usage data, so the switch, Do Not Track and Global Privacy Control do not stop it. Each page of a repository that you open counts one view, and each clone counts one clone.
Maintainers see counts only, never who you are:
- Source stores a hash of you for each day, not your account or your IP address.
- After the day ends, Source deletes the hashes.
- Source keeps the daily counts for one year.
Two parts of Insights come from the usage data: the median reading time and the top countries. For these two parts:
- Maintainers see a median and counts only, never who you are.
- A country with fewer than 3 different visitors is not shown.
- When you turn off Share usage analytics, your visits are not in them.
The other counts of Insights stay separate from the usage data. See See who clones and visits your repository for what maintainers see.
Done
You know what Source records about you, where it keeps the data and for how long. The switch in Settings, Profile turns it off for the browser and the server.