A coordinated intelligence, not a single model.
Solidius AI is being designed as an architecture that combines language, reasoning, memory, tools and specialized agents behind one unified interface — under the user's control.
What Solidius AI is — and what it is not.
Most current AI products expose a single large language model behind a chat interface. Solidius AI takes a different direction. It is being built as a coordination layer that treats language, reasoning, memory, tools and specialized agents as separate components that can be composed for each task.
The goal is not to be the largest model available, but to be the most useful intelligence for a given person or organization — because it can remember relevant context, choose the appropriate capability for the moment, and act only within the boundaries the user has authorized.
This page describes the intended architecture, what currently exists in the private preview, and what remains in development. Nothing on this page is a commitment on timeline.
The vocabulary behind the system.
A shared vocabulary matters when the components are as important as the model itself.
Coordination layer
A software layer that receives a request, decides which capabilities are needed, and orchestrates them into a single response.
Reasoning
The structured, multi-step analysis applied to a task — including planning, decomposition and verification, distinct from raw text generation.
Memory
Optional, user-controlled long-term context that lets the system remember preferences, prior work and relevant facts across sessions.
Tool use
The ability to invoke authorized external services — reading a document, calling an API, executing a search — under explicit permission.
Specialized agent
A domain-focused capability (research, code, legal, operations) that can be selected or combined behind the main interface.
Model routing
Selecting the most appropriate underlying model for each task based on requirements such as depth, latency, cost or locality.
A single request, several coordinated steps.
The diagram below outlines the intended request lifecycle. Individual components exist at different levels of readiness.
- 01
Understand
The request is parsed for intent, constraints and required context. Ambiguity is surfaced rather than guessed at.
- 02
Recall
Relevant elements from user-controlled memory are retrieved. If memory is disabled, this step is skipped entirely.
- 03
Plan
The system decides whether the task needs reasoning, tools, specialized agents, or a combination — and in what order.
- 04
Act
Authorized tools are called or specialized agents are invoked. Any sensitive action requires explicit user approval.
- 05
Verify
Outputs are checked against the request. Uncertainty is stated honestly rather than hidden.
- 06
Respond
The final response is returned with a clear trace of what was used and what was done.
Current capabilities and roadmap.
Solidius AI is in private preview. The lists below separate what exists today from what is still being built, without hiding either.
- Coordinated conversation interfaceA unified chat surface that can receive tasks and return structured responses.
- Bilingual support (English / French)All user-facing content and interactions are available in both languages.
- Model-agnostic routing (early)The orchestration layer can route requests to different underlying providers rather than being tied to one.
- Progressive access controlsEarly-access invitations gated behind an explicit request, not open sign-up.
- Long-term user-controlled memoryConfigurable memory with full transparency and per-item edit and delete.
- Specialized agent orchestrationDomain-focused agents (research, code, operations) coordinated behind the main interface.
- Authorized tool useStructured, permissioned execution of external services with a verifiable action journal.
- Multimodal understandingExtending the interface to documents, images, audio and structured data.
- Developer platformAPI, SDK and connectors for teams building on top of Solidius AI.
How decisions get made.
- 01
Coordination over scale
The value comes from combining the right components, not from operating the largest possible model.
- 02
Honest readiness
A capability is either available, in development, or planned. It is never labelled with language that implies more than it delivers.
- 03
User control by default
The user decides what the system remembers, what it can access, and what it is authorized to do.
- 04
Model independence
The orchestration layer is designed to remain provider-agnostic, so the system can evolve as models evolve.
- 05
Traceability
Any non-trivial action leaves a record the user can review — a black box is not an acceptable design.
Questions people ask about Solidius AI.
Is Solidius AI its own foundation model?
No. Solidius does not currently train or operate a proprietary foundation model. The system is being designed as an orchestration layer that can route tasks to the most appropriate underlying model, which allows it to evolve as the model landscape evolves.
How is this different from a standard chatbot?
A standard chatbot exposes one model behind a text interface. Solidius AI is being built to combine reasoning, memory, tool use and specialized agents. The interface is similar; the architecture behind it is not.
Can I use Solidius AI today?
Solidius AI is in a private preview. Access is granted progressively through the Early Access programme as individual components reach genuine readiness. There is no open sign-up.
What data does Solidius AI keep about me?
Long-term memory is optional and disabled by default. When enabled, remembered items remain visible, editable and deletable. The full data policy will be published as the corresponding components reach public availability.
Which underlying models does Solidius AI use?
The orchestration layer is being designed to remain model-agnostic. The exact set of underlying models will be documented publicly once the routing layer reaches general availability.
Is Solidius AI open source?
The orchestration layer is not open source. Individual components, connectors or reference tooling may be published openly if and when doing so genuinely helps developers integrate with the platform.
How does Solidius AI relate to Solidius.io?
Solidius AI is the intelligence portal of the wider Solidius ecosystem (wallet, digital identity, payments, blockchain infrastructure, professional services). It is being designed to progressively integrate with these components while remaining usable on its own.
Can I build on top of Solidius AI as a developer?
A developer surface — API, SDK, agent tooling, connectors — is planned but not publicly available yet. The Developers page tracks the state of that work.