One governed context engine. Four modules.
Cortx captures the knowledge trapped in your documents and people, governs it with provenance and audit trail, and deploys it to any AI tool or agent. Four modules do that work — extraction, governance, audit, and the workspace where it runs.
Extraction, governance, audit, and the workspace — one stack.
Together they are Context Ware: the layer between AI tools and organisational knowledge, the way firmware sits between hardware and software. Vertically integrated, so context keeps its provenance from capture through to the moment a model reasons over it.
Govern
Holds context in a governed, versioned, auditable format — the .ctx file: definitions, standards, skills, the rules of a domain. Every version is retained; every change is attributable.
Ingest
Pulls structure out of documents and tacit knowledge out of people, at scale — the raw material of context, captured before it walks out the door.
Prove
Embeds chain of custody and audit trail at the moment context is created, not bolted on at retrieval. Governance that runs headless.
Work
The workspace where work happens inside the harness — context-constrained, fully observed, every decision recorded and replayable.
Five governed types. Each with its own rules.
Knowledge is organised into five types, each with distinct deployment rules and permission models. Work is draft-first — nothing auto-publishes.
- Skill — a reusable capability a model or agent can apply.
- Method — a methodology or procedure: how the work is actually done.
- Policy — a rule or constraint that governs what a model is allowed to do.
- Reference — authoritative facts, definitions, and data.
- Package — a set of the above, deployed and versioned as one unit.
Content Managers — the domain experts who own knowledge accuracy — review and approve before anything goes live. Approval is a role, not an afterthought.
Answers from your knowledge — that cite their source.
A Cognitive Twin is a conversational interface grounded in your institutional expertise. It retrieves, reasons over, and cites your organisation's own knowledge — not the internet, not generic training data. Every answer traces back to the governed context it came from, so a reader can check the source, not just trust the output.
Deployed where the work already happens.
Context deploys through the Model Context Protocol to the tools your teams already use, and to the agents you build. One governed source, many consumption points — change the context once and every surface reflects it.
Your knowledge stays under your governance.
- Encrypted at rest and in transit.
- Role-based access with comprehensive audit logging.
- No training on customer data — your knowledge is never used to train third-party models.
- Deploy where you choose; retract when needed.
- Tenant-isolated deployment for sovereignty-sensitive organisations.
See what your organisation already knows.
Cortx is in invite-only beta. Talk to us, or start for free.