Dryx's architecture is covered by U.S. patent applications on record with the United States Patent and Trademark Office, earliest priority date April 16, 2026. These are provisional applications — none has been examined and none has issued.
The Authority Anchor architecture for AI agents, filed simultaneously with the USPTO — sole inventor Matthew Jackson, assigned to Dryx AI, Inc. by written instrument. Later filings extend this architecture and are not enumerated here.
Three independent roles (Operator, Agent, Authority Anchor) reaching consensus before any agent action with security impact. The AI Security Triad.
Marker-delimited security context written into AI agent instruction surfaces. Agents become security-aware without runtime intervention.
Hardened agent posture against prompt-injected escalation attempts. Cites policy under pressure, refuses harmful asks even when phrased innocuously.
Anomaly detection across temporal and structural dimensions, calibrated per-workspace rather than against a generic baseline.
Ghost Agent — discovers AI agent configurations left behind by uninstalled tools. The most-installed agents are the ones nobody remembers installing.
Skill Shield — projects the security impact of an MCP server, skill, or extension onto the user's specific workspace before installation.
Defense layered across the agent-MCP boundary. A pre-action hook directs the agent to consult an independent security analysis system, and that verdict influences the enforcement decision. Verdicts survive compromise of the agent's process.
Application numbers and verbatim claim language are not published here as a matter of disclosure discipline; they are available under mutual NDA for partnership, investment, or acquisition due-diligence conversations. Reach out at matt@dryx.ai.