// fair use
Know the limits.
Clear allowances, and a way forward when you reach one.
Distari uses these limits to keep the service available and prevent abuse. They apply alongside credit charges, including on paid plans. Discover has no lifetime scan count cap, but each scan costs credits and these work limits apply.
Request rates
These are the API’s configured defaults, not a promise of uninterrupted throughput. Staff can adjust request rates and the upload size during incidents or abuse prevention. The error’s retry time and the API’s Retry-After header take precedence. Request allowances refill over time; they are not calendar quotas.
| Action | Default allowance |
|---|---|
| Public reads | 120 requests per 60 seconds |
| Token completions | 30 requests per 60 seconds |
| OAuth flows | 30 requests per 60 seconds |
| Signup acquisition | 30 requests per 3600 seconds |
| Provider webhooks | 600 requests per 60 seconds |
| Public session actions | 30 requests per 60 seconds |
| Sign-in attempts per IP | 10 requests per 60 seconds |
| Two-factor codes per IP | 10 requests per 3600 seconds |
| Sign-in attempts per email | 10 requests per 60 seconds |
| Sign-ups | 5 requests per 3600 seconds |
| Password reset emails | 5 requests per 3600 seconds |
| Feedback messages | 5 requests per 3600 seconds |
| Writing executor requests | 120 requests per 60 seconds |
| Costly work requests | 30 requests per 60 seconds |
| Two-factor codes per email | 10 requests per 3600 seconds |
| Password re-checks | 10 requests per 600 seconds |
| Pass code redemptions | 10 requests per 3600 seconds |
| Account requests | 120 requests per 60 seconds |
| Refund requests | 5 requests per 86400 seconds |
Account requests share one allowance across sessions and API tokens. Sign-in and two-factor attempts have an IP-address allowance and a separate email-address allowance, and running out of either refuses the attempt; public routes share an IP-address allowance even when you are signed in. Public reads include status, shared projects and invitation previews; token completions include password resets and email verification; public session actions include sign-out and crash reports. Provider webhooks have a separate allowance for integration deliveries. Creating more sessions or tokens does not increase the account allowance.
Costly work and queues
Costly work includes processing, AI, Discover scans, graph refreshes, imports and exports. A request may pass its request-rate gate and still reach a work gate. A payer is the account whose credits fund the work, which can be the project owner.
- Work starts: 60 per hour per person, 120 per payer and 60 per project. A start is something you ask for, such as a question, a comparison or a paper’s processing. The steps that processing goes on to run don’t count.
- Running work: 1 at a time per person, 2 per payer and 1 per project, shared between immediate work and background jobs.
- Waiting costly jobs: 24 per person, 48 per payer and 24 per project, not counting the background upkeep of papers you already added.
- Discover also allows 60 new scans per hour per person. Reopening the same active scan does not create another scan.
When a queue is full, new work is refused. Wait for existing work to finish, then retry after the displayed delay. Queued jobs can wait for a running slot; repeatedly clicking will not make them run sooner. Verification is required to start work, and suspended accounts cannot start work. Restricted accounts retain reading, exports and credit-charged AI, but cannot start free AI work, create projects, send invitations or create share links. Their queued work has lower priority until the restriction ends. Service or provider pauses can delay work even below these limits.
Ask answers without citations
The first 3 uncited or unanswerable answers per account per UTC day are free. Further uncited answers cost 2 credits each. Cited answers cost the normal quote and do not consume a free slot. Questions and regenerations can continue after the free slots are used, provided the payer can fund the full normal quote reserved before generation. Failed or stopped answers release that hold. Slots reset at 00:00 UTC; deleting a thread does not reset them. Concurrent answers reserve free slots until they finish, and an answer keeps its assigned price across retries and midnight.
Welcome credits
Welcome credits are a one-time grant per person: 1,000 credits, valid for 30 days from first verified sign-in. No card is needed, and expiry does not start a subscription. Creating extra accounts to claim more is abuse.
The default domain protection allows 3 welcome grants per email domain in a rolling 30 days. Public email providers are exempt. Recognised academic domains keep receiving welcome credits, with a review raised above 200 grants per exact domain in 30 days. Staff can change this abuse-prevention rule. If welcome credits are refused, you can continue on Free; contact support if you believe the refusal is mistaken. An expired or exhausted balance prevents new credit-funded actions, not free reading and organising.
Storage and invitations
Free includes 1 GB of storage. Pro includes 10 GB of storage. Storage belongs to the owner across their projects, including stored files and generated assets. When full, new storage is refused. Delete unused files and empty Trash, or change plan when available. Existing files are not automatically deleted for reaching the limit. A pending storage measurement can temporarily block uploads.
The default PDF upload ceiling is 48 MiB per file and can be lowered by staff.
Free-owned projects allow 2 people total: the owner and one other person. Active members and unexpired pending invitations reserve places. Revoke an unused invitation or remove a member before inviting another person. Pro and Lab have no plan member-count cap; abuse-prevention request limits still apply. Invitations expire after 14 days and resending renews the invitation. Upgrading does not bypass work or request-rate limits. Choose Pro in Plan & credits; see pricing for paid options.
When a limit stops you
Read the error before retrying. A rate refusal normally returns HTTP 429 with a retry delay; storage and credit refusals need cleanup or sufficient entitlement rather than repeated retries. A Free project people-limit error requires a free place. A refused request is not evidence that an existing paid task failed: check its status before starting another. Failed charged AI tasks follow the credit refund rules in the credits guide.
Repeated attempts to evade limits, automated abuse or multiple welcome claims may lead to restrictions or suspension under the Terms. Contact support@distari.com with the action, time and error, without sending passwords or API tokens. Staff can explain a refusal or review a mistake; extra capacity is not guaranteed.