Technology
Intelligence as an operating layer.
Mythos treats AI the way a company treats its operating system: one shared layer with memory, rules and tools that every product runs on. The concept is public. The implementation stays internal.
Mythos OS
Five layers, one direction.
Mythos OS
The orchestration layer. Work enters as a task, is planned, assigned to an agent, reviewed and closed, with a record of what happened.
AI / Agents
Models as interchangeable workers. Requests are classified by the kind of work and routed to the right provider, including a self-hosted open model. Switching a provider never means rewriting what the system knows.
Knowledge / Data
Memory with provenance. Every fact, decision and document keeps its source, its date and its trust level, so what the system knows can be audited and corrected.
Execution
Supervised, sandboxed runs. Isolated workspaces, no secrets in reach, retries, timeouts and a review gate before anything is merged or deployed. Service state is probed and published, never declared.
Products
What reaches the market. SPY, IDauto and AgriBee are built on this layer and share its memory, contracts and rules.
Principles
Three rules the platform is built on.
Provider independence
No knowledge lives only in a provider's prompt. Context and memory are the company's; models are workers.
Read-only by default
Monitors, gateways and integrations start with read-only access. A write path is a deliberate, documented decision.
Evidence over claims
Status is probed, tests are run, deployments are stamped with their commit. If it was not measured, it is shown as unknown.
Knowledge, execution and capital-and-risk infrastructure are proprietary and not published.