arrow_backBack to the overview
The product

One governed knowledge substrate for every person and agent.

Institutional knowledge at the core. Personal knowledge and working memory around it.

The Corvair Knowledge Substrate manages the knowledge an institution depends on: licensed expertise, institutional content and files, and vetted external sources. It makes that knowledge available to authorised humans and agents at parity, grounded in the same policies, provenance, citations, and warrants. Memory preserves the continuity of the work. Vault lets each person supplement the institutional base with private knowledge of their own.

One substrateinstitutional knowledge and files governed once
Human and agent paritythe same capabilities and evidence, through different interfaces
Supporting contextMemory for continuity, Vault for personal knowledge
hub Knowledge Substrate Governed institutional core Institutional inputs Licensed expertise, files Vault, supporting Personal knowledge Memory, supporting Continuity across time person Human Browser, mobile in future smart_toy Agent API and MCP Same authority, same evidence, same governed capability
The core product

The institution's knowledge lives in the substrate.

The Knowledge Substrate is the governed institutional layer. It manages licensed knowledge, institutional files and content, and vetted external sources. Memory and Vault support that core without changing its authority: Memory contributes continuity, while Vault contributes personal knowledge when the user and policy allow it.

Institutional inputs Licensed knowledge Expert content you license Your institution Content, policies, files Vetted external Reliable outside sources Personal knowledge Vault, supporting Each person's own files Private by default Supplements, deliberately The Knowledge Substrate Governed, cited, trust-scored, and pinned to the state used Memory, supporting Continuity across sessions Your people Browser today, mobile in future Your AI agents Governed API and MCP
  1. Institutional inputLicensed knowledgeExpert content you license.
  2. Institutional inputYour institutionContent, policies, and files.
  3. Institutional inputVetted externalReliable outside sources.
  4. Supplements, deliberatelyVault, personal knowledgeEach person's own files, private by default.
  5. The core productThe Knowledge SubstrateGoverned, cited, trust-scored, and pinned to the state used.
  6. ContinuityMemoryCarries decisions, actions, and context across sessions.
  7. Served at parityYour people and your AI agentsPeople through the browser today and mobile in future. Agents through the governed API and MCP.

The Knowledge Substrate remains authoritative for institutional knowledge. Vault supplements it with personal knowledge on a separate, deliberate path. Memory carries context across time. Humans and agents reach the same governed substrate under the same authority.

Core product

Knowledge Substrate

Manages licensed expertise, institutional content and files, and vetted external sources. Personal Vault knowledge may supplement that base under policy. Every answer is grounded, cited, trust-scored, and pinned to the knowledge state used.

Licensed expertise Institutional content Institutional files Vetted external sources
Explore governed knowledge
Supporting subsystem, personal knowledge

Vault

Keeps each person's own files, notes, research, and working artefacts private. It checks and organises them, then lets them supplement the substrate without silently becoming institutional record.

Explore Vault
Supporting subsystem, continuity

Memory

Preserves what was said, decided, changed, assigned, and produced, then turns that record into the working context needed to continue.

Explore Memory
Human and agent parity

The interface changes. The governed capability does not.

People and agents search, ask, browse, frame, create deliverables, and verify answers against the same Knowledge Substrate. They are subject to the same permissions, source boundaries, provenance, and warrant rules.

Parity does not mean every actor sees everything. It means a human and an agent with equivalent authority receive equivalent governed capability and evidence.

Knowledge you can stand behind

Start with knowledge you can stand behind.

The Knowledge Substrate remains the centre of the system. It is where information becomes dependable enough to ground an answer, a deliverable, or an agent action. The sections below explain how Corvair answers, structures, proves, and keeps that knowledge current.

Answer and cite

One query across the governed sources, returning a plain-language answer with the citations behind it.

Frame and deliver

Structured views keep the shape of a complex question and turn it into a deliverable you can reuse.

Verify and prove

A Validity Warrant seals a signed, replayable record of the basis on which an answer was given.

Why it matters

Search gives you a list. A generic AI gives you a guess.

A search box hands you ten links and leaves the real work to you. A general-purpose AI assistant sounds confident but cannot tell you where its answer came from, and it will invent something plausible when it does not actually know. In regulated work, a confident guess you cannot check is a liability, not an answer. Corvair answers only from content you have vetted, shows the exact sources behind every claim, and will not pass off a weak source as a strong one.

list

A search box

Returns links, not answers. You still read, compare, and reconcile it all yourself, and it has no idea which source to believe.

smart_toy

A generic copilot

Answers fluently but cannot show its sources, mixes public guesswork with your data, and fills gaps with things that are simply not true.

verified

The Corvair virtual expert

Answers from your governed knowledge only, cites every source, shows the facts it relied on, and can seal the answer as proof.

How it works

Ask in plain language. Get an answer you can stand behind.

You ask the way you would ask a knowledgeable colleague. The virtual expert replies in plain language, drawn only from the governed knowledge, and every reply arrives with its receipts: the sources it used, the specific facts it rests on, and the option to seal it as a record.

The Ask view: a virtual expert returns a grounded, cited answer with a premise chain and a signed Validity Warrant.

The Ask view in the app: a grounded, cited answer with its premise chain and a signed Validity Warrant.

01 You ask a question In plain language 02 Answered from your base Not free web browsing 03 Cited, with its reasons Sources and premise chain 04 Seal it as proof A Validity Warrant

The same four steps whether a person or an agent is asking. Nothing is answered from ungoverned browsing, and any answer can be sealed for later.

Mode

Search

Survey what the base knows across every source at once.

Mode

Q&A

A direct, cited answer with multi-turn follow up.

Mode

Browse

Follow the map of topics and filter by tag and dimension.

Mode

Deep research

A long run of queries assembled into one cited report.

tips_and_updates
All four modes mix freely in a single conversation. Reach the same virtual expert through a chatbot, search, a secure API, or an MCP server, so it also sits inside the tools your team and your agents already use.
Why the FrameGraph matters

A single answer throws the structure away.

Ask a hard question and a normal assistant gives you one paragraph. The moment you want the whole picture, who is involved, what is at stake, what happened when, and why, you are back to assembling it by hand. The FrameGraph fixes that. It lays your answer out along a framework you choose, such as who, what, when, where, why and how, a SWOT, or a market analysis, over the underlying map of people, places, and events. Instead of a paragraph you get a structured, clickable view you can explore, reuse, and turn straight into a report.

A single answer Read it once, then start over The same answer, framed Your topic Who What When Where Why How Explore it, reuse it, export it as a report

Same question, two outcomes. On the left, a paragraph you consume and discard. On the right, a FrameGraph: a navigable view you can click into, read ten different ways, and hand off as a finished deliverable.

bolt
The second view is nearly free. A framework is just a way of looking at knowledge the base has already gathered, so switching from, say, a 5W1H view to a SWOT reuses the work rather than starting over. What you get is not a throwaway chat reply but a reusable asset your team keeps building on.
Why the Validity Warrant matters

Proof on every answer. Think DocuSign® for decisions.

You can rarely guarantee that a decision is correct. You can always prove the basis it was made on. A Validity Warrant is a sealed, cryptographically signed record of an answer, and it can be reopened and replayed years later for an auditor, a regulator, or a court. That is what turns a governed answer into evidence.

01 The question asked The exact question and any follow-ups 02 What was known The facts used, each with its trust score, at that moment 03 How it was reached The path from question to answer, and why each source 04 Who asked, and the bounds The person or agent, and whether it stayed within policy 05 Signed and replayable Cryptographically signed, timestamped, reconstructable SIGNED & SEALED verifiable later

One warrant, five things recorded. It can seal a single answer, a span, or a whole conversation, and it reopens on any future date so the basis for a decision is never lost.

gavel
When a supervisor asks why a decision was made, you do not reconstruct it from memory and email threads. You open the warrant. A decision cannot be guaranteed correct, but its basis can be proved, which is what makes the substrate safe to rely on where the consequences are real.
Trust you can see

Every fact is rated, and stale facts quietly lose their standing.

Not all sources deserve equal weight, and nothing stays true forever. Corvair scores each fact by how reliable its source is, then lets that confidence fade over time unless the fact is refreshed. New material must clear a quality gate before it is allowed in, so weak or unchecked content never slips into an answer unnoticed.

T1

Institutional, certified

Regulators, boards, and official records. Highest trust.

T2

Credentialed, reviewed

Peer-reviewed and expert sources. High trust.

T3

Self declared, general

Everyday and general sources. Moderate trust.

T4

Speculative

Unverified or speculative. Low, and always flagged.

fact_check
Before anything enters the base it must pass a quality gate covering eight checks, including accuracy, completeness, timeliness, provenance, and corroboration. Anything below the bar is held for a person to review, never admitted silently.
Who uses it, and when

Built for the people who ask, curate, and answer for decisions.

Each person sees only what their role allows, scoped to the knowledge bases they are entitled to. The same governed answer serves everyone, from the analyst asking a question to the auditor checking how it was reached.

person

Business user

Asks, searches, and gets grounded answers within bounds. Can suggest sources and curate their own files.

smart_toy

Virtual expert

The plain-language assistant that answers from the governed base, grounded and carrying a warrant.

account_tree

Steward or curator

Shapes what the base knows: which sources, how far to trust them, and how it stays current.

policy

Compliance & audit

Opens and verifies the Validity Warrant behind any answer or decision.

settings

Administrator

Runs the platform: schedules, retention, and limits. No access to the content itself.

hub

Autonomous agents

Act through the API under an explicit authority matrix, grounded in the same trusted base.

Where it earns its keep
What it delivers

Faster answers, provable decisions, one source for people and machines.

Seconds

to a sourced answer, instead of hunting across drives and inboxes.

Weeks

not quarters for a new joiner to reach competence.

Every answer

source attributed and able to carry a Validity Warrant.

One source

shared by your people and the agents that assist them.

See it, then see how it runs.

Talk it through with us, open the working prototype to use it as a person and a steward would, or read how the platform deploys, secures itself, and stays inside your borders.

All trademarks, including DocuSign®, are the property of their respective rights holders and are used for identification purposes only.