Citation MonsterSubprocessors
Last updated
These are the companies that handle customer data so that Citation Monster can run. Each processes it only to provide the service, under a contract. None may use your data to train AI models. We update this page before adding a new one; the Privacy policy explains the rest.
Subprocessors
Vercel Inc.
- Purpose
- Hosting: runs the website, the app and its background jobs, and stores files (a private store).
- Data
- Everything you put into Citation Monster passes through it; uploaded and generated files are stored there.
- Location
- United States (Washington, D.C. region)
- Terms
- Infrastructure only; does not use customer content to train models.
Neon (Databricks, Inc.)
- Purpose
- Database hosting (Postgres).
- Data
- Account data and your workspace's data: papers' text, references, notes, highlights, chats, research runs, knowledge bases.
- Location
- United States (US East)
- Terms
- Infrastructure only; encrypted at rest.
Upstash, Inc.
- Purpose
- Rate limiting (counters in Redis).
- Data
- Account and workspace identifiers in counter names; no content.
- Location
- United States
- Terms
- Infrastructure only.
Anthropic, PBC
- Purpose
- AI models (Claude): reading papers, extracting and checking references, chat, research runs, knowledge bases. Its web search and fetch tools look up sources while checking references.
- Data
- The paper or the parts of it a task needs, and your question or instruction.
- Location
- United States
- Terms
- Commercial API terms: no training on our customers' inputs or outputs; kept up to 30 days for abuse and safety monitoring, then deleted.
OpenAI, L.L.C.
- Purpose
- AI models (GPT) when you choose one, and web search in chat.
- Data
- The paper or passages a task needs, and your question or instruction.
- Location
- United States
- Terms
- API terms: no training on API data. We turn response storage off on every request (store: false); kept up to 30 days for abuse monitoring, then deleted.
Google LLC
- Purpose
- Embeddings (Google Cloud Vertex AI, gemini-embedding-2): turning passages of your papers and your search questions into vectors, for search within your papers. Google Workspace hosts our support mailboxes (support@ and the other @citationmonster.com addresses).
- Data
- Passages of your papers (with the paper's title and section heading) and the questions you search or ask in chat; emails you send us.
- Location
- United States and other Google Cloud regions (global endpoint)
- Terms
- Google Cloud terms (Vertex AI): Google does not use our data to train or improve its models; prompts may be kept for a limited period only to detect abuse, then deleted. Grounding with Google Search is not used.
Mistral AI SAS
- Purpose
- Optical character recognition (OCR): reading scanned pages.
- Data
- The PDF, or the pages that need OCR.
- Location
- European Union (France)
- Terms
- Paid API terms: no training on our inputs or outputs; kept up to 30 days for abuse monitoring, then deleted.
Resend, Inc.
- Purpose
- Email: sign-in links, invitations and notifications such as 'your analysis is ready'.
- Data
- Your email address, and in notifications a paper's title and a few findings about its references.
- Location
- United States
- Terms
- Delivers email only.
Stripe, Inc.
- Purpose
- Payments and billing.
- Data
- Your billing details. Card numbers go straight to Stripe; we never see or store them.
- Location
- United States
- Terms
- Payment processing only.
Public data sources (not subprocessors)
We look works up in these public scholarly databases. They receive search terms and identifiers such as DOIs and titles, never your files, notes, chats or identity. What they return is public material, which we store and share across users.
- OpenAlex: Works, authors, citations, abstracts
- Crossref: DOIs and publication metadata
- Unpaywall: Where a legal open-access copy is
- Europe PMC and PubMed: Life-science papers and metadata
- OpenCitations: Citation links
- arXiv: Preprints
- doi.org: Resolving DOIs to publisher pages
Services you connect
Sources (connectors) you add, and outside tools you give access to a knowledge base, are your choice and not our subprocessors. They receive what you share with them, under their own terms.