This list reflects the current sub-processors Krux engages to process customer data. We notify Enterprise customers of changes within 30 days of any update. Material changes also publish here before they take effect.
Sub-processor
Purpose
Data processed
Location
DPA
Supabase
Postgres database, authentication, and file storage
Accounts, RFPs, uploaded documents, drafts, cost models, audit log
Per project region (US or EU)
Vercel
Application hosting and serverless compute
Request handling; data in transit and in-memory at compute time
United States (project region)
Anthropic
LLM inference (Claude family)
Prompts and retrieved RFP context at inference time only
United States
Voyage AI
Embeddings (voyage-3) and reranking (rerank-2)
Document and query text chunks at embed and rerank time
United States
Modal
Chemical-structure recognition (DECIMER OCSR) for the chemistry features
Structure images extracted from uploaded RFP PDFs
United States
Mcule
Building-block purchasability lookup (public API)
InChIKey identifiers only (derived structure hashes); no documents
United States
PostHog
Product analytics (usage events)
Usage telemetry. Configured to avoid customer documents and prompts.
Per POSTHOG_HOST configuration