Scan your first repo

Go from a codebase to an editable dataflow diagram in a few minutes. You need Node 20+; you don’t need an account until step 2.

Supported languages: TypeScript, JavaScript, Python, and Terraform (.tf / .tfvars), including common AWS, Azure (azurerm), and Kubernetes (kubernetes_*) provider resources.

1. Run the scan

From the root of the repository you want to scan:


npx @dataparade/cli scan .

The structural scan runs on your machine: your code is never uploaded, .env files are never read, and the output is deterministic — see What leaves your machine for the full guarantees. Structural scans are free and unlimited; they never touch workspace scan quotas.

On success the CLI prints:


[scan] dataflow.json written to /path/to/dataflow.json
[scan] Preview ready: https://app.dataparade.io/preview/cli/<claimToken>

Useful variations:

Monorepo package or subfolder: npx @dataparade/cli scan packages/api
Skip build artifacts or vendored code: npx @dataparade/cli scan . --exclude dist --exclude "**/vendor/**"
All flags: Scan arguments

Open the Preview ready URL from the terminal.
Sign up (magic link or Google) — no account is required to run the scan, but you need one to save the diagram.
After sign-up, the dashboard opens with Preview & Edit loaded from your scan.
Review assets, third parties, actors, and data flows, then create the assessment.

Full details: Upload to dashboard and Creation and Imports.

Alternative: manual import

If you skipped auto-upload or prefer uploading the file yourself:

Sign in to your DataParade workspace.
Open Create assessment on the dashboard and choose CLI import.
Upload dataflow.json.
In Preview & Edit, review the detected content, then create the diagram.

3. Optional: workspace API key (existing users)

Workspace owners can create a CLI access key under Workspace → Access keys. With the key set, the preview link opens the dashboard directly for that workspace (?importDraft=):


DATAPARADE_WORKSPACE_API_KEY=<your key> npx @dataparade/cli scan .

Uploads never consume scan quota. Opt out with --skip-auto-upload. See Upload to dashboard.

4. Optional: AI enrichment

The structural scan is deterministic and never calls a model. If you want AI to fill in third-party metadata and data-flow insights, opt in with your own provider key (BYOK):


SCAN_AI_INFERENCE=true \
SCAN_BYOK_PROVIDER=openai \
SCAN_BYOK_MODEL=gpt-4o-mini \
SCAN_BYOK_API_KEY=sk-... \
npx @dataparade/cli scan . --ai-inference

Every AI suggestion must cite evidence as exact repo-relative file paths and line ranges from your scan — proposals without valid evidence are rejected. See AI inference for providers, scopes, and cost controls.

Something look wrong?

Missing service, backwards flow, empty diagram, or a monorepo the scanner choked on? See Report a scan problem — scan reports are how new frameworks and stacks get supported.