Persistent context
Give your assistant a name and a durable workspace. Conversations, approved memory, documents, projects, tasks, and local state can remain useful across sessions instead of disappearing when a tab closes.
NexxusOS Desktop brings conversation, memory, files, writing, private browsing, Deep Research, coding, integrations, and approval-gated tasks into one persistent desktop workspace.
Built for people who want an AI that can understand their work, stay useful across sessions, and move beyond answers — without hiding what it can see or what it has done.
Windows 10 & 11 · Local memory · Local or API models · Approval gates · Action receipts
Most AI tools begin and end with a conversation. NexxusOS Desktop gives the conversation somewhere to go.
Your agent can work with the files, documents, browser pages, research jobs, codebases, tasks, and integrations inside your workspace. When work becomes consequential, NexxusOS can preview the action, ask for approval, and preserve a receipt of what actually happened.
Give your assistant a name and a durable workspace. Conversations, approved memory, documents, projects, tasks, and local state can remain useful across sessions instead of disappearing when a tab closes.
Move naturally between Chat, Files, Writer, Browser, Deep Research, Codebase, and Tasks without rebuilding the context of your work in a collection of unrelated applications.
NexxusOS separates suggestions, staged actions, approvals, execution, and receipts. You can see what is being requested, what is allowed, what is blocked, and what has actually completed.
NexxusOS is designed to route work by evidence — not simply let a model claim that something happened.
Each NexxusOS surface has a clear responsibility. Together, they form one working environment instead of a loose collection of AI features.
Talk with your named assistant in a familiar conversation interface. NexxusOS can use current attachments, workspace context, and approved local memory while keeping current evidence more important than old assumptions.
Bring files into a local workspace, organize project material, and move directly into a dedicated writing environment for documents, notes, technical content, and drafts.
Use the built-in browser for visible web context or launch a durable Deep Research job for broader investigations. Follow research lanes, sources, findings, progress, reports, and receipts from one place.
Register a local repository, inspect its structure without executing project code, ask focused questions, prepare bounded context, identify tests and project commands, and turn coding objectives into durable governed work.
Turn an objective into durable step-by-step work. Track progress, approvals, blockers, command results, background work, and receipts without relying on a chat message to remain alive.
See what NexxusOS can currently access, which model is active, which sources are configured, what permissions apply, and what remains blocked. Keep local files, browser sessions, API connectors, terminal results, and model providers visibly separate.
NexxusOS is designed to help with real work without turning control into a black box.
Start in Chat with a question, file, project, research request, coding objective, or task.
NexxusOS routes the request to the relevant workspace, current file, visible browser page, codebase, configured integration, or local memory.
When an action could change a file, run a command, use an integration, or affect another system, NexxusOS can stage the work and show what is being requested.
You decide whether governed actions should continue. Missing permissions, unsupported actions, and unsafe requests remain visible instead of being silently ignored.
Completed work can leave a receipt containing the target, decision, result, timestamps, output summary, and available restore information.
A proposed action is not an executed action. NexxusOS is built to preserve that difference.
Investigate a topic across current web sources, local files, visible browser pages, and workspace context. Keep findings and citations connected to a durable report.
Move from conversation to a working document without losing the source material, project context, or decisions that shaped the draft.
Understand a local repository, prepare focused code context, identify tests and commands, review changes, and track coding work through governed verification stages.
Turn recurring or multi-step objectives into Tasks, Work Orders, approval requests, routine drafts, and receipt-backed outcomes.
NexxusOS Desktop stores its user profile, memory, workspace state, documents, and local artifacts under the current Windows user's local application and workspace paths.
Supported local models can run directly on your machine. Web research, provider APIs, and external integrations use a network connection only when you configure and use them. NexxusOS keeps the active model route and connected sources visible, so you can understand where your work is happening.
Your local profile, workspace state, memory, and local artifacts belong to your Windows user instance.
Use supported local models, optional provider APIs, or governed routing policies without replacing the workspace around the model.
A visible browser page is not silently treated as API access. Local Git is not GitHub API access. A model statement is not an action receipt.
Files, browser actions, terminal work, integrations, and automation can be allowed, approval-gated, or blocked by policy.
Trust-significant actions can record what was requested, what was approved, what ran, the result, and whether restoration is supported.
See what NexxusOS can access right now and disable browser control, shell work, external integrations, routine execution, or non-local model routes when necessary.
Local-first means the default center of gravity is your computer — not that NexxusOS disguises connected features as offline ones.
NexxusOS separates the working environment from the reasoning engine. Change the selected model route without discarding your files, tasks, workspace organization, or local continuity.
Run supported models through the managed local runtime when privacy, offline availability, or direct hardware use matters most.
Configure supported provider access when a task benefits from a remote model. Provider use must remain visible and subject to local policy.
Choose local, API, or hybrid routing policies based on the workload while preserving the same NexxusOS workspace and trust controls.
Local-model speed and capacity depend on the user's CPU, RAM, GPU, VRAM, model size, and context settings.
NexxusOS Desktop is an actively developed Windows beta. The current product already brings together its core workspace, while some advanced integrations, coding workflows, automation routes, and device-control features remain bounded, approval-gated, or dependent on configuration.
Features may change during beta, and updates may be required. External services require the appropriate account, connector configuration, scopes, permissions, and network access.
Actively developed Windows beta · Updates included
Windows 10 or Windows 11, 64-bit.
16 GB RAM minimum. 32 GB or more recommended for larger local workloads.
CPU-only operation is supported but slower. An NVIDIA RTX GPU with 8 GB VRAM or more is recommended for a strong local-model experience.
Allow 8–10 GB free (15–25 GB recommended for multiple local models).
Enter your name and email to receive the current Windows beta download.
Download the complete beta package and extract the full folder to a location where you have read and write access.
Keep Setup.exe and the accompanying payload file together, then launch Setup.exe to create your local NexxusOS Desktop instance.
Windows may display additional security prompts for an unsigned beta build.
Get the current Windows beta and create a NexxusOS instance built around your files, your memory, your model choices, and your permission settings.
Bring your conversations, files, research, documents, code, tasks, models, and permissions together in NexxusOS Desktop.