Audience
Who we are targeting.
sift is for teams that treat documents as sensitive infrastructure — not content to upload into someone else’s SaaS.
Ideal customers share one constraint.
The documents cannot casually leave the building. They need grounded chat for people, review for quality, and APIs for agents — on infrastructure they control.
Personas
Knowledge and ops admins
Upload documents, manage collections, and configure processing without wiring three separate tools together.
Reviewers
Legal, compliance, and subject-matter experts who approve, edit, or reject low-confidence extractions before indexing.
Employees
Ask questions across an approved corpus and get answers with citations — not generic model improvisation.
Platform and AI engineers
Connect agents through REST, MCP, and the CLI. Scoped keys, markdown export, and context packs — not a chat-only dead end.
Security and IT
Deploy on-prem or air-gapped, isolate tenants, and keep an audit trail of uploads, reviews, and agent access.
Where this fits
Regulated or privacy-sensitive organizations with large corpora and growing agent workflows.
Legal and compliance
Contract packs, policies, and handbooks that cannot leave the firm’s boundary.
Healthcare and life sciences
Protocols, SOPs, and research corpora that demand isolation and review.
Finance and internal IT
Procedures, runbooks, and regulated knowledge that agents and humans both need.
Research and engineering orgs
Large document corpora where layout-aware extraction and citations matter more than demo chat.
If your constraint is trust — not novelty — sift is aimed at you.