Workspace scan quotas

When you start a hosted scan from the DataParade app — GitHub, GitLab, or a local ZIP upload from the dashboard — the scan draws from your active workspace quota. CLI scans can also draw from the same workspace quota when they use platform AI with a workspace API key. All members of that workspace share the same limits and usage counters.

What never uses quota: structural CLI scans (the default npx @dataparade/cli scan .), CLI scans that bring your own AI provider key (BYOK), and CLI upload / auto-upload are free and unlimited. Quota applies only to hosted scans started from the dashboard and to CLI scans that use platform AI.

This page covers limits, when usage is deducted, and the pre-scan project estimate shown before you start. For team settings and workspace navigation, see Workspace and team. For hosted import steps, see Creation and Imports.

Limits

Each workspace has two lifetime pools (defaults below; your administrator can raise them per workspace):

Notes:

• Limits apply to the active workspace, not per user.
• Structural enrichment (local heuristics without an LLM) does not count toward the platform AI token pool.
• CLI scans with a workspace API key but without AI inference do not call quota preflight/complete and do not consume scan slots or platform AI tokens until/unless you use hosted billing paths.
• Anonymous CLI upload/claim: claiming the preview after signup consumes 1 scan slot (and any platform AI tokens accrued during an anonymous AI session). The Anthropic/platform key never leaves the server — the CLI only receives a short-lived session token for the infer proxy.

Quota deductions

Rules differ slightly by how you run the scan:

What is deducted

• Completed scan slots: +1 only after a successful hosted job or CLI run that used the quota API and finished without a failed completion report.
• Platform AI tokens (hosted): actual LLM usage from the completed job’s AI inference metadata, not the pre-start estimate.
• Platform AI tokens (CLI): each platform infer call during the scan; see Activity for CLI completion lines that note tokens were billed during infer tasks.

What is not deducted

• A scan that failed (upload error, clone failure, scan error, CLI exit before success, and similar).
• A start blocked by quota at preflight or pre-start (no job consumption).
• Invalid input (bad ZIP, missing repo, and similar) — may appear in Activity but do not consume a slot or tokens.
• CLI structural-only scans with a workspace key (--no-ai-inference / SCAN_AI_INFERENCE=false) — no preflight, no slot, no platform tokens.

Where you see usage update

• Workspace → Usage — progress bars for completed scans and platform AI tokens. Use Refresh or revisit the page after a scan completes.
• Workspace → Activity — scan.started, scan.completed, scan.failed, and scan.quota_blocked. CLI completions may show Source: CLI scan and whether tokens were billed during infer tasks.

Passing the pre-start check does not reserve tokens. Hosted scans bill tokens on completion; CLI platform scans bill tokens as infer tasks run.

CLI scans with a workspace API key

When AI inference is on (the default) and you set DATAPARADE_WORKSPACE_API_KEY (or --workspace-api-key):

1. Preflight — POST /api/scans/cli/preflight checks scan slots and estimated AI budget from project footprint (fileCount, bytesIngested).
2. Platform infer — LLM calls go through the DataParade API (POST /api/scans/cli/ai/infer/tasks, submit + poll).
3. Complete — POST /api/scans/cli/:jobId/complete records success or failure.

If preflight fails because quota is exhausted, the CLI prints:

[scan] workspace quota: No scan slots remaining in this workspace.

(or a similar message for insufficient platform AI tokens). That is a quota block, not a scan crash.

With a workspace key but AI inference off, the CLI runs locally only — no preflight, no completion report, no quota usage. See Scan arguments and Environment variables.

Pre-scan project calculation (hosted dashboard)

Before you click Create on a Git or ZIP hosted scan, the app checks whether the project is likely to fit your remaining budget. This is a heuristic estimate, not a full dry-run of the scanner.

The CLI performs a similar footprint check at preflight when platform AI and a workspace key are enabled.

When the estimate runs

While the check runs, you may see Checking token budget… in the create dialog.

What is calculated

1. Project footprint — approximate size of what will be scanned:
   • Git: file count and ingested bytes from the GitHub or GitLab API.
   • ZIP: derived from the compressed archive size (approximate unpacked footprint).
2. Estimated token need — a rough minimum–maximum range from the footprint (files, megabytes, estimated AI candidate count, and typical provider-call cost). The range is shown in blocking messages when the scan cannot start.
3. Per-job budget — how many tokens this scan may draw: remaining pool capped at 400,000 per job.
4. Allow or block — if the estimated minimum tokens exceed the per-job budget, or you have no scans remaining, Create is disabled and a notice explains why. Examples:
   • This scan needs ~12,000–45,000 tokens (842 files, ~18.3 MB). You only have 8,000 left.
   • This scan needs ~25,000 tokens (120 files, ~4.1 MB). Only 20,000 can be used for this scan.

Important caveats

• The estimate is approximate. Actual token use on completion may be lower or higher.
• A passing estimate does not reserve tokens for that scan.
• If fewer than about 1,000 platform AI tokens remain in the pool, hosted AI scans cannot start (platform minimum).

UI touchpoints

When you are blocked

Related topics

• Workspace and team — Settings, Usage, Activity, and Teams
• Creation and Imports — hosted Git and ZIP flows
• CLI Run and Export — flags, env vars, and platform AI
• Troubleshooting — quota blocks for hosted and CLI scans