# Files and privacy

> Know what stays local and what can change in your project.

Web page: https://supercov.com/docs/workspace-isolation
All Supercov pages for agents: https://supercov.com/llms.txt · Full text: https://supercov.com/llms-full.txt

Supercov runs locally. It doesn't require an account or upload your source,
tests or coverage reports to a Supercov service. Your coding agent's own data
handling is separate; check its settings and provider policy.

## What changes in your project

Supercov stores reports, execution data and assertion assessments under `.supercov/`,
which it excludes from Git. Your agent's test edits are ordinary repository
changes that you can review, commit or discard.

Supercov measures an instrumented copy of the project.
It does not overwrite your source files with instrumented code.

The test command still has its normal side effects. A test can write a file,
update a snapshot, call a service or modify a database. Supercov is not a
sandbox for arbitrary commands; use the same precautions as an ordinary test run.

For JavaScript, TypeScript and Rust, new or modified command outputs in the
isolated workspace are copied back, with exceptions for instrumented source
and deletions. A change a formatter or fixer in your test command makes to a
source file (`prettier --write`, `eslint --fix`) reaches your project when the
command ends. Python and Ruby run in place, so test-command file changes and
deletions happen directly in your project.

## Run suites at the same time

You can run several Supercov commands in one project at once, such as unit
and end-to-end suites in parallel CI steps. Each run works in its own copy of
the project, so none waits for another. Combine them afterwards with `merge`:

```sh
npx supercov -- <unit test command> &
npx supercov -- <end-to-end test command> &
wait
npx supercov merge
```

With no run named, `merge` combines the runs recorded on the current code.
Extra copies are removed a day after their last run, or with
`npx supercov runs clean`.

To add new tests without running the whole suite again, measure only the new
tests and run `merge`. Helpers and test runner settings changed with them keep
an earlier run when its tests don't load them, or when the change only adds
functions or a timeout. The [CLI guide](https://supercov.com/docs/cli.md#combine-shards) lists what
counts.

## What uses the network

The first npx supercov invocation may contact the npm registry to download the package.

Your tests and coding agent may also use the network. Supercov's local storage
doesn't change how those tools connect to external services.

## Keep or remove reports

Run reports let the agent compare results and revisit earlier work. They may
include file paths, source excerpts and assertion details, so treat them as
project data when sharing an archive or filing an issue.

Long-running tasks can accumulate reports and build caches. Ask the agent to
show what cleanup would remove before approving it, and keep any runs you still
need for a comparison. See [Speed and storage](https://supercov.com/docs/performance.md) for cleanup
options and [Runs and evidence](https://supercov.com/docs/evidence.md) for the stored files.