Claude Code plugin

To use Qualflare from Claude Code, install the Qualflare plugin and run /qf-init once in your project; then /qf-cover, /qf-run and /qf-fix write tests, upload results with qf <identifier> collect and fix failures from the chat.

1. /qf-cover generates tests for the source files you changed, or for a whole file or directory with --all, and writes them after your approval.
2. /qf-run runs your test suite and uploads the results to Qualflare, or uploads a report you produced yourself.
3. /qf-fix reads the last run's failures, makes the smallest source fix and re-runs once.

An optional coverage nudge reminds you, after each response, when you edited source files but no tests.

Before you begin

You need:

• Claude Code, the CLI or desktop app.
• The Qualflare CLI (qf) on your PATH; see Installation.
• Node.js on your PATH, for the coverage nudge: its hook runs node.
• A Qualflare project and a project access token. Creating a token: Editor+ · project (requires the Editor role or higher in the project). Uploading results doesn't need Allow writes. See Access tokens.

Let your coding agent install qf

Optional: paste this prompt into Claude Code to install qf for you.

Install the Qualflare CLI (`qf`) on my machine.

1. Detect my operating system (macOS, Linux, or Windows) and CPU architecture
   (amd64 or arm64).
2. If Homebrew is available (macOS/Linux), run: brew install qualflare/tap/qf
3. Otherwise, open https://github.com/qualflare/qualflare-cli/releases/latest,
   download the asset matching my OS and architecture
   (qf_<version>_<os>_<arch>.tar.gz, or .zip on Windows), extract it, and move
   the `qf` binary into a directory on my PATH.
4. Run `qf version` to confirm the install succeeded.

Install the plugin

/plugin marketplace add Qualflare/qualflare-claude-code

/plugin install qualflare@qualflare

The first command adds the Qualflare marketplace. The second installs the qualflare plugin from it, and the plugin stays installed in later sessions. Claude Code's plugin documentation explains where a plugin can be installed.

You should see the /qf-* commands in Claude Code.

Set up your project

/qf-init

qf login <identifier>

Create tokens in your project's Project Settings > Access Tokens (https://app.qualflare.com/project/<project-identifier>/settings/access-tokens). The identifier is only a local name for the CLI, so it doesn't have to match the project. Uploading results doesn't need Allow writes. In a monorepo, each package gets its own identifier — run qf login once per package.

Generate, run and fix tests

Once setup is done, three commands cover the typical cycle.

Generate tests for what you changed

After writing or editing source files:

/qf-cover

/qf-cover reads .qualflare/test-state.md to know your frameworks, finds the source files changed in this session that lack tests, proposes 2 to 5 test cases per function, and writes them after your approval. Pass a path or glob to scope it:

/qf-cover src/auth/login.ts
/qf-cover packages/api/src/**

By default /qf-cover only covers code you changed this session. Pass --all to cover an entire file or directory regardless of recent changes:

/qf-cover --all src/utils/

With --all, or in a project with no tests yet, it asks you to narrow the scope to 8 files or fewer per round.

Run the suite and upload results

When you're ready to verify:

/qf-run

/qf-run runs your test suite and uploads the results to Qualflare via qf <identifier> collect. Pass a framework slug, such as /qf-run jest, or a package path, optionally followed by a slug, such as /qf-run packages/web jest, to run part of the suite. To upload a report you produced yourself, pass its path and slug: /qf-run zap-report.json zap. For a directory the slug is required, and without one /qf-run asks. For a file without a slug, it detects the format with qf validate and asks you to confirm. Cypress results upload as Mochawesome JSON, so add mochawesome as a dev dependency first. Uploads go to the environment named in QF_ENVIRONMENT, or to development when it is unset. Qualflare refuses an environment the project doesn't have.

/qf-run jest
/qf-run playwright

See Environments for QF_ENVIRONMENT, and the /qf-run reference for the frameworks it runs itself.

Fix failing tests

When tests are red:

/qf-fix

/qf-fix reads the last run's result files in .qualflare/results/ and makes the smallest source change for each failing test. It edits source files, not tests: it skips a test whose assertion looks wrong. It then re-runs the affected frameworks once and reports what still fails; run /qf-fix again for another pass. It doesn't upload, so run /qf-run to send the new results to Qualflare. Pass a slug or path to scope which failures to fix:

/qf-fix jest
/qf-fix packages/api

Turn on the coverage nudge

When the nudge is on, the plugin checks the session each time Claude finishes a response. If you edited a source file and no test file, and at least one of those edits was substantive, it prints:

🔍 Qualflare: 3 source file(s) changed without test updates. Run /qf-cover to add coverage.

It appears after each response until you edit a test file or turn it off with /qf-hook off. Config files, type definitions and trivial edits don't count. Turn it on or off at any time:

/qf-hook on
/qf-hook off

Or check the current setting with /qf-hook (no argument).

Keep the state current

Inspect current state:

/qf-state

Shows detected frameworks, file counts, hook setting, and which CLI identifiers are configured locally.

Refresh counts after adding tests:

/qf-update

Faster than a full re-init — re-globs each framework row and updates .qualflare/test-state.md in place. Pass a slug or path to scope the refresh.

Update the plugin

Pull the latest version at any time:

/plugin update qualflare

Check that it works

• Run /qf-state: each identifier shows configured.
• After /qf-run, the summary ends with Uploaded to Qualflare ✅, and the launch appears in your project (Analyze launch results).

Troubleshoot

Run /qf-doctor first when something feels wrong:

/qf-doctor

It checks the qf CLI, each package's login, the slug list, config.json and the hook, the age of test-state.md, file-count drift and your framework tooling. It reports ✅, ⚠️ or ❌ per check with a fix command; /qf-doctor lists the checks and thresholds.

When an upload fails, /qf-run prints the CLI's error. Upload errors explains each exit code, and CLI troubleshooting covers the CLI itself.

How the plugin differs from Quo and MCP

The Qualflare plugin runs in Claude Code, where you write code, and drives the qf CLI on your machine to write tests, run them and upload results. Quo is the agent inside the Qualflare web app: it works on the project's data, analyzing coverage, suggesting cases and generating steps. The MCP server gives any MCP client, Claude Code included, a live connection to the project's data; the plugin uses no MCP connection.

Related

• Guides: Quo agent · Connect a coding agent (MCP)
• Reference: Claude Code plugin reference · Collect command
• CLI and API: Installation · Configuration · Access tokens