Privacy policy

Last updated 15 October 2026. This page describes how SourceLab behaves today. It is not legal advice, and a lawyer has not reviewed it. Text in brackets is unpublished information.

Who publishes SourceLab

SourceLab is published by Harsh. A registered legal name, postal address, and governing law are not stated here: [Legal entity name], [Postal address], [Governing law]. The public site address is [Production domain]. This text says “this website” until that domain is confirmed.

There is no published support email yet: [Support email]. Until that address exists, SourceLab has no in-product contact form.

Accounts

You can create an account with email and password, or with Google. Email sign-up asks you to verify the address. Password reset sends a link by email when email delivery is configured. The account stores your name, email address, and, for Google sign-in, the profile image Google provides. The password itself is not stored in plain text by the application database.

Sign-in uses an httpOnly session cookie (SameSite=Lax, and Secure when the server is in production). Signing out ends that session in this browser. There is no button to delete an account.

Notebooks, sources, and chat

Notebooks are private to the account that created them. SourceLab does not offer shared or team workspaces. A notebook can hold pasted text, markdown, a PDF, a website URL, or a YouTube URL.

  • Pasted text and markdown are stored in the application database.
  • PDFs are uploaded to Cloudinary. Extracted text is stored in the database. Deleting a source removes the database row and its search index. It does not delete the Cloudinary file.
  • Website import sends the URL to Firecrawl and stores the resulting text.
  • YouTube import fetches the transcript and stores that text.
  • A web-search result can be saved as a source. Chat can also send a query to Tavily when web search is allowed for that turn.

Ready sources are split into chunks, embedded, and stored in Pinecone under that notebook. Chat messages, citations, and a rolling summary are stored in the database. Prompts, source text, and model output are sent to OpenAI to embed content, answer questions, and generate study tools. SourceLab does not control how long OpenAI keeps that data.

Memory

Memory is stored with Mem0. User memory is recalled in every notebook you own. Notebook memory stays in the notebook where it was saved or learned from chat. You can add, edit, and delete memories from the memory page. Deleting a notebook asks Mem0 to delete that notebook’s memories. It does not delete user memory. If Mem0 is not configured, memory features do nothing.

Credits and payments

Free includes 10 credits once. Pro is ₹499 per month and sets the balance to 500 credits each billing period. Unused Pro credits do not roll over. Chat costs 0.1 credit, a learning artifact costs 1, and a source that finishes processing costs 1. Stripe processes checkout and the billing portal. SourceLab stores the plan, credit balance, and Stripe customer id. Card numbers are handled by Stripe, not by SourceLab’s database.

Cancelling Pro keeps the Pro plan until Stripe ends the subscription. Credits left at that point stay on the account. No refund policy is offered.

Analytics

Google Tag Manager is not loaded until you allow Analytics, and only if a container id is configured for this deployment. Advertising storage stays denied. The choice is kept in this browser until you change it. Details are on the cookie page.

Deletion and retention

Deleting a notebook removes its database records, and SourceLab attempts to delete that notebook’s Pinecone namespace and notebook memories. Those outside deletes are best-effort. Deleting a source removes its database text and Pinecone vectors, not the Cloudinary PDF. User memory, the account, Stripe customer data, and copies held by OpenAI, Mem0, Cloudinary, Resend, or other processors are not cleared by those actions.

A retention period has not been set: [Data retention period]. There is no self-serve account deletion. A deletion request cannot be completed from the product until a support address is published.

Other processors

Depending on the feature, data can be processed by PostgreSQL, Pinecone, OpenAI, Cloudinary, Firecrawl, the YouTube transcript service, Tavily, Mem0, Resend, Stripe, Inngest, and Google (for sign-in, and for Tag Manager if you allow Analytics). Each of those services has its own retention rules. SourceLab does not claim a certification or a data-processing agreement on this page.