Full-stack private AI
AI that knows your company.
Workers that do its work.
Nothing leaves your perimeter.
Archaea builds the private AI stack enterprises deploy on their own servers. Biblio turns scattered documents, databases, and code into precise context AI can answer from. ArcFact puts AI workers on the clock — planning, executing, and reporting, 24 hours a day.
SELF-HOSTED · MODEL-AGNOSTIC · BUILT FOR REGULATED INDUSTRIES
Two products
The knowledge layer and the worker layer
They aren't competitors — they're two layers of the same stack. Each stands alone in a deal; together they turn AI workers into employees who actually know your company.
Enterprise RAG system for AI agents
Biblio turns a company's scattered documents, database tables, code, and third-party system data into clean context AI can actually use — retrieval augmentation plus a knowledge graph, all inside your perimeter.
- Six retrieval modes: local, global, mix, graph, vector, bypass
- flow-lite parsing engine: PDF, Office, images, table extraction
- Knowledge graph + vector store over your entire corpus
- Seven built-in datasource connectors, plugin-extensible
- Headless API plus a web console for knowledge librarians
Your AI. On the clock. Around the clock.
ArcFact is an AI worker platform — a virtual factory of digital employees that runs 7×24 on your own servers, taking repeat operations, analytics, compliance, and document work off your team's plate.
- AI workers that plan tasks and collaborate in parallel
- Drag-and-drop Blueprint workflows — no code required
- Scheduler for around-the-clock, audit-ready execution
- Three-layer sandbox isolation, down to the kernel
- Open ecosystem: MCP, embedded browser, terminal, Skills
One stack
Shared core, flexible coupling
Every ArcFact deployment ships with Biblio's flow-lite parsing engine inside — the same document intelligence, everywhere. The full Biblio platform attaches when your corpus demands precise, pre-indexed semantic retrieval.
flow-lite inside, always
ArcFact embeds Biblio's flow-lite engine for OCR, deep document decomposition, table extraction, and semantic understanding. Even a standalone ArcFact deal ships Archaea's core document technology under the hood.
Attach Biblio when data grows
Workers reach data on demand through MCP, the embedded browser, and the terminal. Once you've accumulated tens of thousands of documents, the full Biblio platform gives workers precise semantic retrieval over a pre-indexed corpus.
Bring any model
The LLM layer is your call: a fine-tuned agentic model, mainstream open-source models on your own hardware, or a MaaS endpoint. All three paths plug into the same architecture and can be mixed or swapped as needs change.
Why Archaea
Built for the environments others can't survive
Archaea are the oldest organisms on Earth — they thrive where nothing else can. We build AI the same way: engineered for sealed, regulated, compliance-heavy environments where data cannot leave the building.
Data sovereignty
Full private install on your servers. No SaaS dependency, no data off-premises — a shortlist requirement in finance, healthcare, manufacturing, and the public sector.
Isolation to the kernel
Three-layer isolation — row-level database, filesystem, and kernel-level sandboxing — with full audit logs, RBAC, and SSO integration.
Our own document engine
The flow-lite parsing engine is Archaea technology, not a bolted-on library — built for the complex PDFs, Office files, and tables real enterprises run on.
Platform plus delivery
We don't stop at a license. Group-scale engagements rebuild your most painful workflow end to end — platform, customization, and long-term operations.
The company
Who is Archaea?
Archaea AI, Inc. is a Silicon Valley company that owns and develops the technology behind Biblio and ArcFact. We go to market worldwide through authorized resellers — in Japan, our products are sold by Krugle Inc. — with additional markets opening across Asia-Pacific.
We turn your scattered company knowledge into precise context your AI can answer from — then put AI workers on it, around the clock. None of it leaves your perimeter. The Archaea thesis
See the stack on your own data
A proof of concept starts with your documents and your worst repeat workflow. Short cycle, low risk, inside your perimeter from day one.