Build
Author governed agents in plain English or code — skills, knowledge, guardrails, and a no-code studio with 11 role packs.
Governed agentic AI
The operating system for governed AI agents.
Build, govern, observe, and distribute AI agents across regulated enterprises — with 21,500+ governed skills and agents ready to install, every run gated, every decision provable.
Aligned with EU AI Act · NIST AI RMF · ISO 42001
Control Tower
Fraud TriageExecuted
Payout ApproverHold
Return ReviewerExecuted
KYC ScreenerBlocked
trust-ledger ▸ 1,284,902 actions recorded today
Built for regulated industries — financial services, healthcare, and the public sector
NORTHWINDMERIDIAN BANKATLAS HEALTHCIVIC AGENCYHELIOS ENERGYVANTAGE TRUST
Illustrative sectors — not customer logos.
See it end to end
Four minutes, the whole operating system.
One governed run, from the prompt to the signed receipt — the gates, the ledger, the budget caps, and the console your team works in.
4 min 11 secProduct footage from the Cortex console — no reenactments.Watch on YouTube ↗
The library
21,500+ governed skills & agents.
24 industries. 12 roles per industry. 18 agent patterns. Every item installs from the catalog, wires to your systems, and runs through the same fail-closed gates as everything else on Cortex — 21,511 items today, counted live from the catalog.
The problem
Shadow agents are a governance time bomb.
Teams are scaling AI past pilots with no audit trail, no policy gate, and no kill switch — exactly the exposure EU AI Act and NIST AI RMF were written for. Cortex puts every agent under one governed runtime.
Ungoverned sprawl
no identity · no audit · no caps
Governed runtime
identity → gates → ledger
One runtime, four jobs
Everything between the prompt and production.
Govern
Every run passes the same gates: identity, budget, guardrails, registry, control tower, execute, output guard, audit. Fail-closed by design.
Observe
Monitor, score, evaluate, simulate, and debug every run — with reliability gates and a tamper-evident trust ledger underneath.
Distribute
Ship a governed solution once and deploy anywhere as signed, verifiable Solution Packs — across tenants and air-gapped environments.
The platform, end to end
Every control between prompt and production.
40+ governed capabilities across one runtime — the full operating system for enterprise AI, not a scattering of point tools.
Build
- No-code Agent Studio
- Behavior & guardrails
- Knowledge & RAG
- Agent marketplace
Govern
- Policy-as-Code
- Action Fabric
- Agent IAM
- Oversight modes
- Risk register
- MCP gateway
Automate
- Workflows
- Automations (ECA)
- Channel intake
- Batch runner
- Real-time hub
- Schedules & queue
Observe
- Monitoring
- AI Insights
- Evaluation
- Simulation lab
- Reliability score
- Agent debugger
Data & Ontology
- Business ontology
- Object & property permissions
- Asset catalog
- Lineage & provenance
Cost & Value
- Cost governance & caps
- Adoption & ROI
- Business value
- Model ops & compute broker
Distribute
- Solution studio
- Signed Solution Packs
- Pack registry
- Certification
How it works
Follow one run from author to proof.
Scroll the lifecycle: the visual on the right tracks each step a governed run actually takes.
studio ▸ build → publish
Author
Build an agent with skills, knowledge, and plain-English guardrails — or install one of 21,000+ skills and agents from the library. Publishing requires a passing reliability score.
8 gates · deny is the default
Every run hits the gates
Identity → budget → guardrails → registry → control tower → execute → output guard → audit. Over budget returns a 402, out of policy a 403, conflicts a 409 — before anything touches your systems.
HOLD · human approval
High-risk actions pend
Payouts, refunds, POs, access grants — anything crossing your thresholds stops on an approval board with full context. The agent proposes; your people dispose.
hashOk: true
Every outcome is provable
Executed, held, or blocked — the run seals into a tamper-evident, hash-chained ledger with 10-hop provenance. Replay it, hand it to an auditor, verify it offline.
Agent Studio
skills
Search KBCreate ticketSend email add
Publishedv4 · reliability 96
Where is order #4471?
Shipped, arriving Thursday.tools: Search KB · $0.004
studio ▸ build → publish → chat-test
Action Fabric
Dry-run›
Refund $240
risk: low · evidence
Proposed›
Close case #4471
risk: med · evidence
Pending approval›
Payout $5,000
risk: high · evidence
Executed
Email customer
risk: low · evidence
actions ▸ dry-run → propose → approve → execute → compensate
audit/verify
ok: true
verifyChain ▸ chained SHA-256 · signed receipts
97%Tax pack accuracy SLA (published target)
95%Citation-coverage SLA on grounded answers
10-hopProvenance on every outcome
100Live-verified run quality score
Prove it — don't just claim it
Every number traces to its source, in a ledger you can hand an auditor.
Cortex records every run in a tamper-evident, hash-chained Trust Ledger — with signed receipts you can verify offline and a 10-hop provenance graph from human to outcome.
- Tamper-evident, hash-chained audit (verifyChain detects any edit)
- Signed outcome receipts, verifiable offline
- Datapoint provenance: every fact links to its source filing, page, and box
audit/verify
ok: true
verifyChain ▸ chained SHA-256 · signed receipts
Solution Packs
One signed pack shipping today — two reference blueprints.
Versioned bundles of ontology, policies, actions and agents — install into your tenant, verified by hash and signature. The Tax pack ships signed today; Claims and KYC/AML are reference blueprints on the roadmap.
Security & compliance
Built for the enterprise security review.
Fail-closed gates, RBAC, DLP, agent detection & response, and multi-tenant isolation — mapped to the frameworks your auditors already use.
SOC 2ISO 27001ISO 42001HIPAAGDPREU AI ActNIST AI RMFFINRA
“Cortex is the first thing that let us put agents in front of regulated work. Every decision has a receipt, and I can pause the whole fleet from one screen.”
Run agents. Keep control.
Put your AI agents under air-traffic control — governed, observable, and provable from day one.