deXI_001, a 33-billion-parameter reasoning model, and deXI_0a1, a smaller action model, running side by side without disk swapping.
Xenryx Enterprise / Edge-first intelligence
Project
Argonath.
Privacy-preserving enterprise intelligence. A substrate that understands your organization and carries work through to execution, within the boundaries you define.
From workstation to organizationProvision. Govern. Execute.01 / Foundation
Intelligence that understands the work. Execution that stays accountable.
Project Argonath combines organizational context with deterministic action on workstations. Its edge-first architecture brings reasoning, screen-aware agents, and a live view of orchestration into the enterprise perimeter.
The following baseline is reported in the Xenryx Labs strategy and deployment memorandum, prepared by Sonik Bhandari, Founder & CEO.
A standard hardware server completing more than 49,605 consecutive secured records without interruption or memory loss.
Agents read the screen and carry out repeatable steps in applications and terminals without relying on external API connectors.
Xenryx's visual engine maps multi-agent work in the browser, with a reported rendering baseline of 60 frames per second.
02 / Deployment trajectory
Start at the edge.
Scale on your terms.
Modular Intelligence Packages
Securely provision specific intelligence capabilities into local Xenryx environments. Packages adapt to workstation hardware and the organization's subscription tier, giving each department the intelligence its workflows require.
Zero-retention cloud bridging
The planned Model Context Protocol (MCP) bridge and Stateless Gateway Protocol route eligible reasoning to cloud services. Client-side policy engines protect context before egress and govern what may leave the device.
Isolation for every task
The roadmap places autonomous work in temporary, hardware-virtualized containers with explicit network limits and an auditable execution boundary.
03 / Enterprise applications
Expanding enterprise solutions.
Deployment targets from the Argonath roadmap. Security, regulatory requirements, and offline capabilities are evaluated for each implementation.
Accounting & Finance
Local audit trails, tamper-evident records, and tax-ledger matching.
Inventory & Supply Chain
Edge-first stock and shipment tracking, warehouse forecasting, and local execution.
Defense & Intelligence
Air-gapped deployment, coordinated security agents, and hardware-bound encryption.
Healthcare
Client-side privacy controls and private clinical analysis through governed routing.
Legal & Compliance
Citation-backed local research, isolated privileged documents, and contract checks.
Energy & Utilities
Isolated industrial systems, grid forecasting, and offline operational workflows.
Manufacturing
Live production maps, sensor-based early warnings, and on-site execution.
Telecom & Networks
Private machine-to-machine coordination, edge inspection, and zero-trust routing.
HR & Employee Data
Encrypted employee records, local benchmarking, and governed personal data.
Insurance
In-house risk modeling, claim checks, and traceable policy histories.
Government
Offline citizen records, identity checks, and controlled inter-agency exchange.
Pharma & Research
On-site modeling, traceable trial records, and protected intellectual property.
Construction
Coordinate project documents, site planning, procurement, and field workflows within governed operational boundaries.
Food & Service
Connect inventory, service operations, scheduling, and quality workflows through private, adaptive intelligence.
04 / Continuous Autonomous Loop
Read. Reason. Act.
Keep the loop governed.
Argonath's Continuous Autonomous Loop connects context, decisions, and physical execution inside the company perimeter. Local knowledge lives in the Geometric Vault; HARP makes orchestration visible.
> audit_ledger --target VDR_Q3 --enforce 17a-4
Inspect a target ledger and apply the workflow's retention and audit policy.
Illustrative command syntax from the deployment memorandum.Xenryx Enterprise