Public materials show a Janus-to-Kairos naming/domain transition: the site and legal docs reference both withjanus.com and withkairos.co / Janus AI, Inc.
SourcePublic company, workplace, funding, and market signals
Updated Jul 29, 2026
Kairos (also operating as Janus / withjanus.com) is a San Francisco-based applied AI company building a simulation testing platform for AI agents. The product generates synthetic users and runs multi-turn simulations to surface hallucinations, rule violations, tool-call failures, and other reliability issues before deployment.
Primary product
Janus AI simulation testing platform
Founded
2025
Headquarters
San Francisco, CA, USA
Team size
2-10 employees
Industry
Technology, Information and Internet
Sub-industry
AI simulation testing / AI evaluation
Offices
Business model
Stage
seed
Total raised
$500K
Latest round
Seed · Jun 2025
Latest amount
$500K
Jun 2025 · Y Combinator
Investors
Reliability-first, privacy and sovereignty focused, and oriented toward enterprise and regulated-industry workflows. The public materials emphasize operator expertise, first-party data, and customer-owned intelligence.
Pricing
Demo-led enterprise pricing with usage-based eval credits; no public self-serve price sheet found.
Differentiators
Technology
Customers
Competitors
Public materials show a Janus-to-Kairos naming/domain transition: the site and legal docs reference both withjanus.com and withkairos.co / Janus AI, Inc.
SourceFondo · Feb 2026
Launch coverage summarizing Janus’s positioning, founders, and simulation-driven approach for AI-agent QA.
Product Hunt · Jun 2025
Launch page for Janus describing the product, the simulation approach, and the team behind it.
HireTop · May 2025
Explains the product as a simulation testing platform for conversational AI, emphasizing multi-turn simulations, natural-language rules, hallucination and bias detection, and CI/CD integration.
Y Combinator on LinkedIn · May 2025
Launch post describing Janus as a human-simulation platform for testing AI agents, with realistic conversations, custom rules, and detection of hallucinations, bias, and tool-call failures.