Inferred, not filed
Documents arrive in a flat library. The Brain infers classes, projects, tasks, and obligations from the content itself — there is no folder taxonomy to invent or maintain.
The premise
Most tools file documents away. docOS reads them.
A folder is a guess about where something belongs, made once, by hand, and rarely revisited. docOS takes the opposite stance: drop documents in flat, and the Knowledge Brain infers the projects, tasks, obligations, and current context they imply — backed by evidence, citation by citation, and rebuilt as the corpus changes.
How it works
A compiler for your corpus
The Brain Compiler runs in the background in five disciplined stages. It is the only thing that writes the Brain — everything else reads from it, so the corpus never drifts under a noisy model.
1
Detect
Inventory the corpus and split documents into addressable units — pages, sections, emails, transcripts.
2
Extract
Pull entities, facts, obligations, and roles from each unit, anchored back to its exact source.
3
Build
Merge the evidence into one canonical representation of everything the corpus knows.
4
Analyze
Surface contradictions, missing metadata, and obligations that are still unresolved.
5
Project
Emit the Brain: a document graph, a work graph of projects and tasks, and cited context packs.
Why it's different
Three commitments, built in
Not features layered on a file store — the shape of the system itself.
Governance built in stages
Activate versioned policies only after an affected-item simulation. Required metadata, lifecycle, retention, legal holds, disposition review, and correlated durable audit then enforce together; drafts change nothing.
Verifiable AI
Every answer carries citations to its source. Tools are permission-scoped, every call is audited, and accuracy is measured continuously against golden sets.
One workspace
Six surfaces over one Brain
Every surface is a different view of the same compiled knowledge — so what you read, search, and chat with always agrees.
01
Now
Your current context, inferred — active work, deadlines, and what to look at next.
02
Documents
A flat library with indexed text, classification, intake state, versions, and exact-source delivery.
03
Projects
Inferred and confirmed projects, tasks, issues, and obligations.
04
Reader
Truthfully labelled text or normalized reading views, with versions and verified original download.
05
Brain Ask
Ask in plain language with sources, scope, limitations, and durable history.
06
Brain
Navigate the knowledge graph — documents, entities, and how they connect.
Built on docOS
Everything an app needs, already standardized
The platform underneath the products: a cloud-first foundation every app inherits, so each team builds what makes their product different — not the plumbing every product shares.
Identity & accounts
Users, roles, passkeys, sessions.
Wallet & billing
Ledger, payment intake, subscriptions.
AI assistant
Provider-abstracted, metered, governed.
Admin operations
Authorization, audit, operations.
Governance & audit
Versioned policy simulation, retention, holds, guarded disposition.
Cloud sync
Cloud-canonical, optional local mirror.
Trust
Answers you can trace
AI that you can put in front of regulated work has to show its sources and respect its limits. docOS treats verifiability as a requirement, not a setting.
Citation-backed
Grounded answers show available source coordinates; missing evidence stays explicit.
Permission-scoped
Row-level security — the model never sees what you cannot.
Durable audit
Publication, corrections, runs, and model usage retain durable evidence.
Human gate
Partial Brain publication and destructive product actions require explicit confirmation.