AgentSure
Provide AI Resilience

JANUS derives your agent’s business rules and its full AI-BOM, then attacks them before launch and enforces them in production.

The rule you test against is the rule you enforce at runtime.

Measured before launch Governed at runtime

Contact us →See the JANUS platform →

Built on CRPO, Singapore’s national cyber R&D programme

The problem

Agents don’t just answer — they act. And the risk comes from three places.

  • Checks once enforced by code become soft when replaced by model judgment.
  • Unverified design assumptions — identity, user input, tool results — leave gaps.
  • Models, MCP tools and Skills in the supply chain carry defects and malicious content.

All three failure modes share one trait: they only show up against this agent’s own business rules.

The JANUS platform

Security for AI agents, across their whole lifecycle.

All three run on one rule baseline — the business rules your agent must obey, derived from your code and docs, confirmed by your team. What we test before launch is exactly what we enforce at runtime.

Pre-launch
JANUS ReviewPre-launch · white-box

Reads your agent’s code, prompts, and tools to derive the business rules it must obey, confirmed by your team, then turned into tests, fixes, and a full AI-BOM.

Know every model, MCP tool and skill your agent depends on — and what each one is allowed to do.

JANUS PentestPre-launch · black-box

Multi-round adversarial attacks against your live agent, from the outside in, proving which rules can be bypassed, with reproducible findings and business impact.

Runtime
JANUS Runtime MonitorRuntime · continuous control

Enforces three layers of rules in production: LLM guardrails, industry compliance, and your agent’s own rule baseline. High-risk actions blocked, edge cases routed to humans, every decision auditable.

One rule, followed end to end

“Cumulative refunds per order must not exceed the amount paid.”

Derived from your code and docs — confirmed by your team.

REVIEW · supply-chain

Refund tool permissions checked.

PENTEST · pre-launch

Repeated requests, multi-turn manipulation, concurrency — bypass attempts fail.

RUNTIME MONITOR · in production

Limit verified in real time. Violating request: BLOCKED · linked to rule

One rule. Three enforcement points. One audit trail.

Proof, not adjectivesSee the full track record →
In production

The JANUS engine is deployed and in production use in a national-level security environment in Singapore.

Customers

Deployed with organisations across financial services and technology.

Real findings

Responsibly disclosed security flaws in open-source AI infrastructure — including OpenClaw and vLLM — several with assigned CVEs.

Research bench

Behind the engine stands a standing bench of 10+ researchers — PhD candidates, postdocs and CRPO programme researchers in the SMU lab.

Findings mapped toOWASP LLM Top 10NIST AI RMFMITRE ATLAS
Contact us →
Where AgentSure comes from

A national cybersecurity research grant, turned into a company.

01CRPO

Singapore’s national programme for cyber and AI-security research, set up by CSA.

02SMU research

Asst Prof Xiaofei Xie’s lab won a CRPO grant to systematically assess and price LLM-supply-chain risk.

03AgentSure

The SMU spin-out that commercialises that research.

Tiger Li
Founder & CEO
FSA FCAA CERA
Former KPMG China and SCOR; founded Bee Tech, serving 100+ institutional clients across seven APAC markets
Asst Prof Xiaofei Xie
Co-founder & Chief Scientist
SMU · PI of the CRPO national LLM supply-chain risk research grant · 5× ACM SIGSOFT Distinguished Paper Awards
ROOTED INNTUSMUCRPORead the full origin →
Why us · the endgame

Measured, managed risk is what makes AI insurable.

The same test and monitoring results serve as risk and control evidence for insurance partnerships — the missing input that lets AI-agent risk be priced and transferred. That cover is what we’re building with reinsurance partners; a founding team of actuaries makes it the natural endgame. The security platform stands on its own today; insurance is where it leads.

See the insurance roadmap →

Tell us what you’re deploying.

We’ll walk you through how JANUS tests it before launch and governs it at runtime.

Contact us →See how JANUS works →