Frequently asked
questions.

Elyceum AI is a multi-agent intelligence platform that embeds into your application via API or MCP. It powers your product with agents that model emotional context, maintain persistent memory across all sessions, and learn continuously from real interactions. That makes it fundamentally different from a standard LLM integration or chatbot wrapper.

Standard AI APIs provide one-shot language generation with no state, no memory, and no adaptation. Elyceum is a persistent system: it knows your users across every session, models emotional state to calibrate responses, maintains a full indexed history, and learns from real interactions over time. The gap between a static LLM call and Elyceum is the gap between a search engine and a knowledgeable colleague.

Artificial Emotional Intelligence is our name for what this system is, and the line we draw against artificial general intelligence. It works in two directions. Outward, agents read the emotional state of the person they're interacting with (from real signals in language, pattern, and history) and adjust tone, approach, and response to fit. Inward, and more fundamentally, the system carries its own internal affective signals: functional emotions, not felt ones, that drive its learning. They decide what it attends to, what it keeps, and what shapes it over time. So emotion isn't only something it reads in you; it's the mechanism by which it gets better. We chose the term deliberately, to stand apart from AGI: we're not claiming a general mind that rivals humans across every task. We haven't built that and don't claim to. We're claiming something narrower and real: an architecture where emotion, both interpreted and internal, is the foundation for how the system learns and behaves. It's new and powerful precisely because it isn't a claim about AGI. And to be clear: functional analogs of emotion are not consciousness or sentience.

No. Elyceum is a layer that adds persistent memory, emotional calibration, and learning to your AI interactions. It works alongside existing language models rather than replacing them. Think of it as the cognitive architecture that gives your AI backbone: memory, emotional intelligence, and the ability to learn, layers that raw LLMs don't provide.

Elyceum connects via REST API and MCP (Model Context Protocol), running as the intelligence layer inside your existing infrastructure. Your users interact through your own interface; Elyceum is the cognitive core underneath. A separate management portal lets your team configure agent personas, roles, and behavioral parameters without touching code.

MCP is an emerging protocol standard for connecting AI agents to applications and data sources. Supporting MCP means Elyceum integrates cleanly with tools and workflows that have already adopted the standard, reducing integration complexity and enabling richer data flow between your systems and the intelligence layer.

Through the Elyceum management portal, your team defines agent personas and their behavioral mandates: the scope of what each agent handles, its tone parameters, domain-specific knowledge, and escalation rules. A patient-facing clinical agent has very different parameters from a sales advisor or a compliance analyst. Configuration is applied without code changes to your application.

Initial API integration is typically measured in days, not months. The bulk of implementation time is persona and mandate configuration: defining how each agent should behave in your domain. We work closely with early access partners to make this process as efficient as possible, and will discuss specific timelines during onboarding conversations.

Elyceum learns from real interactions in your environment, not from retraining a model, but from accumulated understanding of your domain, your users, and what works. The system builds richer context over time: which types of interactions resolve well, which user signals indicate particular needs, which approaches land in your specific domain. This is pattern-level learning that improves response quality and efficiency without requiring any intervention from your team.

As Elyceum builds understanding of your domain and users, it makes smarter decisions about when to apply expensive reasoning versus rely on established patterns. Novel or complex situations warrant deep reasoning; familiar patterns with clear precedent can be handled more efficiently. The result is that cost-per-interaction tends to decrease over time as the proportion of interactions requiring heavy computation decreases. We're transparent about this: it's a tendency, not a guarantee, and rate of improvement depends on interaction volume and variety.

Learning is fully isolated per tenant. Your organization's learned patterns, user context, and domain understanding never inform another customer's system, and theirs never influences yours. Each organization's intelligence instance is entirely separate. This is a deliberate architectural decision, not just a configuration choice.

Yes. Each organization runs as a fully isolated tenant: your memory, your context, and your learned patterns are never shared across customers. Data handling, retention policy, and residency requirements are discussed during onboarding and matched to your compliance requirements.

Yes. For organizations in regulated industries or with strict data residency requirements, on-premise and private cloud deployment options are available. This is discussed during the early access onboarding process. Regulated industries (healthcare, financial services) are a core focus and we're experienced with their compliance requirements.

Memory storage is encrypted at rest and in transit. Organizations retain full control over what data the system indexes; you can define inclusion and exclusion policies at configuration time. Right-to-erasure and data deletion capabilities are available for compliance with GDPR, HIPAA, and equivalent frameworks. Specific data governance requirements are scoped during onboarding.

Elyceum AI is in early access. We're working with a select cohort of enterprise partners to refine the product and validate it across industry contexts before a broader launch. Request early access and we'll be in touch to discuss fit and timeline.

Pricing is usage-based, structured around interaction volume with the expectation that effective cost per interaction decreases as the system learns your domain. We don't publish a rate card at this stage; early access pricing is discussed directly as part of onboarding, scoped to your expected usage and deployment model. There are no fixed seats or user licences.

We're looking for organizations in domains where context genuinely matters: healthcare, customer operations, financial services, R&D, sales, and HR. The best early access partners are those who have already felt the limitation of stateless AI: they've tried standard integrations, seen the ceiling, and are ready for something that actually learns. If that's you, get in touch.