The Problem
Every session starts from zero. JoshOps AI is where the work is kept.
You explain your project. You correct the same mistake for the fourth time. You close the window and all of it disappears. The model was never the bottleneck — the forgetting was. JoshOps AI is designed to keep that work on hardware you control, preserve sources and corrections across sessions, and require evidence before completion is claimed.
Proposed capability levels
- Aeon — Start: one place for a conversation, approved devices, and the model the user selects.
- Ysolde — Remember: searchable history, source links, and evidence for accepted outcomes.
- Noiriel — Coordinate: multiple qualified models contributing through one accountable path.
- The Hall — Operate: scheduling, governed support, and reusable roles with explicit boundaries.
These are cumulative candidate promises. No plan is available for purchase until a signed entitlement and release evidence prove the promise.
Pricing
- Aeon — $0 candidate early-access tier; not available for purchase.
- Ysolde — $20 per month candidate early-access tier; not available for purchase.
- Noiriel — $100 per month candidate early-access tier; not available for purchase.
- The Hall — $200 per month candidate early-access tier; not available for purchase.
- Nonprofits — JoshOps Impact sponsored-access programs. First pilot: Career Continuity, in review.
Team and Enterprise are contact-sales deployment options built around one of the same four entitlement tiers. They are not additional tiers, and there is no public checkout yet.
How JoshOps AI works
JoshOps AI is designed to keep work continuous across models and machines without handing any model ownership of the user's memory or authority.
A governed request follows seven visible steps: capture the exact request and source; bound the context and effects; select qualified capacity; authorize the exact action; execute through one path; witness the result independently; and close with a terminal receipt and versioned memory.
The operating rules: Customer work stays under customer control. Identity is not authority. Original sources remain separate from derived material. A worker cannot grade its own work. No completion claim is accepted without terminal evidence.
Your work belongs to you
JoshOps AI does not sell customer content, train on it, or use it to target, profile, score, or segment a person. The user retains their rights to their prompts, files, memory, corrections, and derived work.
When you explicitly direct a supported turn to a provider account you choose, that provider receives the selected request and admitted context under its own terms. That user-directed processing does not give JoshOps AI rights to your customer content, and JoshOps AI cannot promise how an independent provider retains or uses data.
Non-content operational telemetry such as crash counts, latency, and version adoption may describe the software. It must not be derived from what you think, want, or work on.
The Architect — Joshua G. Juarez
AI Systems Architect and creator of JoshOps AI. Self-taught, non-traditional. He designs systems that keep original sources separate from derived summaries and require evidence before completion is claimed. His foundation is in customer-facing field operations, owning restaurant technology outcomes across 43 active sites under P1-P4 SLAs.
Early access
Aeon is the $0 candidate tier. JoshOps AI is in early access by request; no tier, installer, account signup, or checkout is publicly available today.
LOCAL_ONLY may cross your private tailnet between machines you control, but no third-party inference provider receives the turn. FLAGSHIP uses one supported connected provider only when you explicitly select it; the request and admitted context needed for that turn go to that provider under its terms. ALL_HANDS may use multiple named admitted lanes; reconciliation remains local. Cloud consent resets after every turn, and a failed cloud request never becomes a silent local success.
hello@joshops.ai