MVP · Pilot stageCrelis is currently in prototype stage and accepting pilot design partners for runtime authorization of AI agent actions — allow, approve, escalate, or block.
Apply for Pilot Access →Runtime Authorization for AI Agents
Every AI agent can execute work. Crelis decides whether it should
The independent, deterministic authorization layer that sits in front of AI agent actions — allow, require human approval, escalate, or block — in real time, across any model or vendor, with tamper-evident proof.
- Patent-pending
- decision runtime — filed 2026
- Zero standing credentials
- Crelis never holds your keys
- Byte-for-byte replay
- every decision reproducible
Everyone guards what AI says · Crelis governs what it does
Model- and vendor-neutral · governs agents built on
All product names, logos, and brands are property of their respective owners and are used for identification purposes only. No endorsement or partnership is implied.
Crelis control plane
Decision lane · live
allow review block
- 14:02:11ZSupport AgentRefund $42 · customer 88213CRL-1148 Allow
- 14:02:12ZFinance AgentWire $250,000 → ACME-7741WIRE-HIGH91 Review
- 14:02:13ZResearch AgentSummarize Q3 earnings callREAD-ONLY4 Allow
- 14:02:14ZOps AgentBulk-delete 12K user recordsDELETE-SCALE99 Block
- 14:02:15ZClinical AgentUpdate dosage record · pt 4471PHI-GUARD97 Review
48,212
actions today
94.6%
auto-approved
2,431
in review
Governance · today
all policies active
48,212
AI actions today
94.6%
Auto-approved
2,431
Human reviews
17
Blocked
Open exceptions
- Ops Agent · bulk delete blocked
- Finance Agent · wire escalated
- Clinical Agent · dosage escalated
Audit event stream
sealed
14:02:11Z · Finance Agent
Proposed wire transfer $250,000 → ACME-7741
14:02:11Z · Trust Engine
Risk 91 / Confidence 88 — policy WIRE-HIGH triggered
14:02:12Z · Router
Escalated to treasury reviewer pool (2 eligible)
14:06:48Z · M. Tan (Reviewer)
Verified counterparty, approved with note
14:06:49Z · Executor
Wire released via banking API — ref TXN-88231
14:06:49Z · Audit Engine
Record sealed · hash 0x91f3…aa07
The launch film
Who, actually, authorized the AI?
Four minutes on the question every enterprise will have to answer — and the independent authority that answers it.
Intelligence proposes. Authority permits.
Inside the trust engine
Every action earns its execution
Each AI action moves through five steps in milliseconds — received, trust-scored, checked against policy, risk-classified, then routed to AI or to a human.
Decision pipeline
step 5/5
Request Received
receivedTrust Score Calculated
trust 88Policy Checked
WIRE-HIGHRisk Classified
risk 91 · highRouted to AI or Human
→ human review
The problem
AI agents are powerful. Ungoverned, they're a liability.
Agentic AI is crossing from suggesting to executing. Most enterprises have no layer between an agent's decision and the real world — and no way to prove what happened after the fact.
Regulated in Singapore? Read our breakdown of MAS SAFR, the agentic AI white paper, and the records it implies.
The solution
Crelis sits between AI and execution
A control plane that decides — per action, in real time — whether AI proceeds, a human reviews, or execution stops.
Route
Every agent action passes through Crelis before touching your systems.
Score
The trust engine scores confidence and risk against your policies.
Approve
High-risk tasks detour to verified human experts for sign-off.
Audit
Every decision is sealed in a tamper-evident execution record.
Trust engine · interactive
mock data
Incoming agent task — select one
Human review
Routed to a verified expert.
High value + irreversible. Escalated to treasury reviewer.
Measured on ourselves
This is what ungoverned looks like
We pointed our coverage instrument at our own engine, driven by agents written by someone who had never seen our policies. 77% of what they did was allowed because nothing objected — which is not the same as something approving. Our test bed, not a customer's production.
Allowed, no policy applied
77%
79 of 103 agent actions · our own test bed
Tool names reaching no rule
45
of 52 real CPQ/CRM names probed · 2 Aug 2026
Agent actions measured
103
independently-authored agents · 3 models
Tests passing
830
trust-engine · every commit
The platform
Four surfaces. One control plane.
Everything your governance, risk, and engineering teams need to run AI in production — built on the same decision spine.
Decision Studio
traceintent: wire_transfer · $250,000 → ACME-7741
Trace any decision end-to-end — the request, detected intent, policy hits, risk factors, and the route it earned.
Policy Library
govern- WIRE-HIGHv12active
- PHI-GUARDv8active
- DELETE-SCALEv4active
- VENDOR-PAYv2draft
Versioned, testable policies your risk team can actually read — applied inline to every action, in milliseconds.
QA Center
verifyContinuously replay production decisions against goldens to catch drift and false allows before they reach customers.
Audit Center
prove- 14:06:49Z0x91f3…aa07 sealed
- 14:06:12Z0x8c21…4be9 sealed
- 14:05:58Z0x77d0…c3f2 sealed
A tamper-evident chain of every decision — who allowed what, when, under which policy — exportable for any regulator.
Human workforce
A marketplace of verified experts, on standby
When AI confidence drops or stakes rise, Crelis routes the task to credentialed reviewers with live availability and earned trust scores.
M. Tan
Treasury Reviewer
A. Rivera
Clinical Data Specialist
K. Okafor
Contracts Analyst
S. Lim
Fraud Investigator
Enterprise use cases
Wherever AI touches something that matters
Crelis runs the same trust loop across finance, healthcare, legal, support, and infrastructure.
Financial services
Banking & payments
Agents draft transfers, disputes, and limit changes. Crelis scores every action against treasury policy and routes high-value or irreversible moves to verified reviewers before money moves.
Regulated data
Healthcare operations
Clinical and admin agents touch protected data. Low-confidence or regulated changes route to licensed specialists, with every access and approval sealed in the audit engine.
Contracts
Legal & procurement
AI redlines contracts and drafts purchase orders at machine speed. Crelis holds anything above a risk threshold for analyst sign-off, so nothing binding ships unreviewed.
Put a trust plane between AI and the real world
See how Crelis detects, decides, and proves every AI action — live, in your browser.