Grounding
- Grounded answers with citationsEvery response links back to the exact source passage.
- Your IP stays yoursYour knowledge never trains a public model.
Uthereal Cortex — The Private AI Agent Platform for Enterprises
The private AI agent platform for enterprises.
The self-serve platform that lets your team turn enterprise data into your own private Agents and powered by your organisational Skills, grounded in your own sources, deployed under your control, no engineering required.
See how it works: sources, agent, skills, deployHow do we structure a market-entry engagement?
Run it in three phases: diagnostic, option design and launch planning. Close each phase with a partner review before the client steering meeting.
- 1Engagement playbook.pdfp. 12
Market-entry engagements run in three phases: diagnostic, option design and launch planning.
- 2Delivery standards.docx§ 3.2
Every phase closes with a partner review before the client steering meeting.
How it works
From your sources to a working agent in minutes.
Four steps, all inside the Cortex workspace. No engineering required.
01 · Ingest
Add your knowledge
Upload your files, then define their tags and purpose. Your material stays in your workspace and never trains a public model.
02 · Ground
Create your agent
Give it a role, a function, a persona and the outcome you want. It answers only from the knowledge you approved.
03 · Automate
Create skills
Turn your expertise into repeatable skills that deliver quality output every time.
04 · Deploy
Publish and share
Preview, then share your agent with your team or customers, and monetise it with Cortex.
The launch film
Watch the four steps end to end.
The same journey: create an agent, add knowledge, create skills, publish.
Two kinds of agent
Knowledge to answer. Skills to act.
Both live in one self-serve workspace your domain experts can drive themselves.
Answer from your sources, with the passage to prove it.
A Knowledge Agent answers questions grounded in your private sources and cites the exact passages it used, so your team can verify and share with confidence.
Knowledge vs Skills Agents in depth- 1Question from your team
- 2Search approved sources
- 3Answer with citations
What Cortex reads
Bring the knowledge you already have.
If you have a source that is not listed, connectors can be added on request.
- Documents
- PDFWordPowerPointExcelCSV
- Knowledge systems
- WikisSharePointNotionWebsites
- Storage
- Google DriveAmazon S3Google Cloud Storage
- Databases
- Structured databases
- Audio & video
- TranscriptsAudioVideo
Platform
Everything you need to run private AI at work.
Governance
- SSO, RBAC, audit logsSingle sign-on, role-based access and full audit trails ship on day one.
- EU data residencyKeep your data in the EU. GDPR compliant. SOC 2 in progress.
Model choice
- Choice of modelsUse the best foundation model for the job and swap without rebuilding your agents.
Delivery
- Skills Agents for workflowsCodify multi-step expert work so anyone on your team can run it.
- White-label and embedShip agents in your own product, site or app under your brand.
- Self-serve, no engineersYour domain experts build and shape agents themselves.
The numbers
Grounded, accurate and yours.
- 25%
- More accurate than generic AI
- >99%
- Extraction accuracy on your content
- 0%
- Of your data given away
- 50,000
- Users already on our agents
Figures as presented in the Uthereal Cortex launch film. Accuracy is measured against generic AI.
Uthereal Cortex vs the alternatives.
A quick side-by-side of Cortex against the enterprise AI options teams usually shortlist.
Yes Built inPartial Limited or add-onNo Not offered
Grounded in your private data with citations
- Uthereal Cortex
- Yes
- ChatGPT Enterprise
- Partial
- Microsoft Copilot
- Partial
- Glean
- Yes
- Custom RAG build
- Yes
EU data residency / sovereignty
- Uthereal Cortex
- Yes
- ChatGPT Enterprise
- Partial
- Microsoft Copilot
- Partial
- Glean
- Partial
- Custom RAG build
- Partial
Your IP never trains a public model
- Uthereal Cortex
- Yes
- ChatGPT Enterprise
- Yes
- Microsoft Copilot
- Yes
- Glean
- Yes
- Custom RAG build
- Yes
Self-serve — no engineers required
- Uthereal Cortex
- Yes
- ChatGPT Enterprise
- Yes
- Microsoft Copilot
- Yes
- Glean
- Partial
- Custom RAG build
- No
Skills Agents (multi-step workflows)
- Uthereal Cortex
- Yes
- ChatGPT Enterprise
- Partial
- Microsoft Copilot
- Partial
- Glean
- No
- Custom RAG build
- Partial
White-label & embed in your product
- Uthereal Cortex
- Yes
- ChatGPT Enterprise
- No
- Microsoft Copilot
- No
- Glean
- Partial
- Custom RAG build
- Yes
Choice of underlying foundation model
- Uthereal Cortex
- Yes
- ChatGPT Enterprise
- No
- Microsoft Copilot
- No
- Glean
- Partial
- Custom RAG build
- Yes
Time to first agent
- Uthereal Cortex
- Minutes
- ChatGPT Enterprise
- Minutes
- Microsoft Copilot
- Weeks
- Glean
- Weeks
- Custom RAG build
- Months
Comparison based on publicly available product documentation as of October 2026. Competitors evolve — verify current capabilities before purchasing.
Ways to use Cortex
Start self-serve. Grow from there.
For your team · Start here
Cortex workspace
Build Knowledge and Skills Agents yourself, with no engineering required. Free during early access.
Start using Cortex for FreeFor developers
Cortex SDK & API
Embed agents deeply into your own products when you need custom integration.
Explore the SDKFor your customers
White-label & custom
Ship agents under your brand, in your product, site or app.
Book a walkthroughFAQ
Questions teams ask before they buy.
Already using Cortex? Setup and troubleshooting answers are in the second group.
Uthereal Cortex is a self-serve enterprise AI platform that turns your private knowledge into grounded, citation-backed AI agents. It combines Knowledge Agents (that answer questions from your sources) and Skills Agents (that run multi-step expert workflows), all under your control with enterprise-grade security.
ChatGPT Enterprise and Copilot are general-purpose assistants layered on top of a shared model. Cortex is built to be grounded in your specific proprietary knowledge, with citations on every answer, EU data residency, and the ability to package your methodology as private white-labelled agents your team or customers can use. Your IP stays yours and never trains a public model.
A Knowledge Agent answers questions grounded in your private sources and cites the exact passages it used. A Skills Agent runs a multi-step expert workflow — think of it as codifying how your best people do a piece of work (a diligence review, a client brief, a research synthesis) so anyone in your team can trigger it consistently.
Your data stays in your Cortex workspace with EU data residency available. It is never used to train public foundation models. Cortex is GDPR compliant and we support enterprise controls like SSO, role-based access, and audit logs. SOC 2 is in progress.
Documents (PDF, Word, PowerPoint, Excel), wikis, SharePoint, Notion, Google Drive, cloud storage (S3, GCS), structured databases, transcripts, and audio/video content. If you have a source that is not listed, connectors can be added on request.
Yes. Every answer from a Knowledge Agent links back to the exact document and passage it was grounded in, so your team can verify and share with confidence.
Consultancies operationalising and monetising methodology, publishers turning archives into Subject Matter Experts, and enterprise teams building a private knowledge base and internal AI on top of it.
No. Cortex is designed as a self-serve workspace that your domain experts can drive themselves — from ingesting sources to shaping agents to deploying them. Engineering is only needed if you want to embed agents deeply into custom products via our API.
Cortex is currently in early access and is free during the pilot phase for qualifying teams. Book a walkthrough to get access and we will scope a pilot for your team.
Cortex is GDPR compliant with EU data residency available. SOC 2 is in progress. The platform ships with SSO, role-based access, and audit trails by default.
A Cortex agent is an AI assistant configured for a specific job. You define what it should help with, connect knowledge sources, customize how it responds, and deploy it when it is ready for users.
Use Create agent, then complete the setup flow: define the agent's name and purpose, add knowledge sources, and review the chat experience before finishing setup.
Draft means the agent is still being configured or cannot be deployed yet. Draft agents may be missing required details, knowledge sources, or completed setup stages.
Yes. You can manage an existing agent to adjust its purpose, knowledge, behavior, chat experience, branding, and sharing configuration without creating a new agent.
You can add supported files such as PDFs, DOCX, XLSX, CSVs, and similar document formats. You can also import websites and connected sources when those integrations are available in the workspace.
Cortex processes uploaded and imported sources before it can reliably use their contents. During processing, the source may appear in the product but answers may not yet include all of that knowledge.
Yes. You can continue configuring the agent while processing runs in the background. Deploy only when the knowledge you expect the agent to use is available.
Tag suggestions appear after a supported source has been added and the backend can generate labels for it. They are source-based, not generated before a file or website exists.
Yes. Preview lets owners test the agent experience before making it live for the intended audience.
Reference visibility controls whether users can see the source behind an answer. Shown reveals source context; Hidden keeps answers clean without inline source markers.
PDF references can open with highlighted evidence. Website references open the related URL. CSV and spreadsheet-style references may be presented as structured table context. Other files may show relevant text excerpts.
The chat may still work with sources that are already available, but Cortex may warn that knowledge base updates are incomplete until all documents finish processing.
Draft agents are not deployed. Live agents have been deployed and are available according to their sharing and access settings.
Yes. If an agent is made available to a real audience, even an internal one, it should be treated as deployed and managed like a Live agent.
Access depends on the sharing configuration. An agent can be limited to approved users, shared by link, or made available through the workspace settings supported by the product.
No. Saving edits updates the agent configuration. Deployment remains a separate action so owners can review changes before making an agent available.
Dashboard summaries and the Analytics page show usage information such as messages, conversations, and performance when real data is available.
That means users have not started conversations with that agent yet, or the usage data has not been recorded for the selected view.
Review the agent purpose, confirm sources finished processing, add clearer knowledge, check reference visibility, and preview the agent before sharing it more widely.
Use Support in Account settings to contact the team with the issue, priority, and any context that helps reproduce the behavior.
Ready to build your own agents?
Join the teams turning their hard-won expertise into private, enterprise-grade AI on a platform their own experts can drive.
Cortex is in early access and free during the pilot phase for qualifying teams.
- 1Sign up and build your first agent for free
- 2Book a walkthrough with the team
- 3We scope a pilot for your team
Explore Enterprise Agents · Sovereign AI · comparisons · features · FAQ