Cross-tool execution
Run one instruction across local files, browser sessions, and selected desktop workflows.
Private by design
Private AI for your computer.
Give one instruction. Velyn plans visible steps across local files, browser sessions, and selected desktop workflows — then acts only with your approval.
macOS private alpha · local Ollama · approval-gated tools
Early access requests open in a BuildLayer form, prefilled for the Velyn private alpha.
What it does
One operator for local files, browser sessions, and selected desktop workflows.
Velyn links a single instruction to concrete actions across local files, browser sessions, and selected desktop workflows while keeping execution visible.
Give intent, inspect planned steps, and approve execution with full override.
Core capabilities
Execution designed for real desktop workflows.
Action plans before execution
Review explicit step-by-step actions before execution begins.
Context-aware assistance
Keeps track of active tasks, open artifacts, and prior direction in the current workflow.
Workflow-level automation
Helps with repeated operating tasks while preserving execution context.
Operator override at every step
Pause, revise, or stop execution at any point in the same thread.
Built for technical work
Designed for codebases, multi-step flows, and tool-heavy daily work.
Privacy and control
Private defaults with explicit control at run time.
Velyn is built so sensitive context stays on your Mac and key execution steps remain inspectable in Operations.
- Local Ollama for AI by default — not a hosted chat brain
- Action plan preview before execution starts
- Permission-aware handling of file and system operations
- Network tools and browsing use HTTPS only when you approve them
- Pause, revise, or stop an active task at any point
How it works
Three steps from intent to controlled execution.
01
Set the outcome
Describe what should be done in plain language.
02
Review the plan
Validate proposed actions across tools before execution starts.
03
Run with control
Approve execution, guide progress, or interrupt when needed.
Use cases
For technical professionals who work across multiple tools.
Turn local notes and managed-browser research into reviewable draft summaries.
Apply file updates, run checks, and keep changes reviewable in one flow.
Handle repeated setup and operations tasks with less context switching.
Gather inputs from tabs, docs, and app output into a structured draft.
Alpha scope
What the private alpha includes today.
Velyn is early software. These boundaries set expectations before you request access.
macOS private alpha
Packaged for Mac (Apple Silicon). Features, stability, and tool coverage evolve during early access.
Local Ollama required
Normal use runs AI on your Mac via Ollama. Setup guides you through installation; cloud LLM providers are not part of the alpha path.
Managed Chromium browser
Web work runs in a separate managed Chromium profile—not your daily Safari or Chrome session.
Selected desktop workflows and allowlisted files
File tools respect allowlisted roots. Workflow actions use curated aliases, not control of every installed application.
FAQ
Common questions about Velyn AI.
Plain-language answers aligned with the current macOS private alpha.
- Is Velyn fully offline?
- No. AI inference runs locally with Ollama, but approved network tools (web search, fetch, browsing) use HTTPS on the internet. That is separate from sending your chat to a hosted AI service by default.
- Does Velyn send my data to the internet?
- Your conversation is not sent to a generic cloud chat brain for normal use. When you approve network tools or browsing, those actions use HTTPS like any desktop tool would. You can review what ran in Operations.
- Which apps and browsers does it control?
- Web tasks use managed Chromium. Desktop app launch is limited to selected aliases in policy—not arbitrary control of every app. LinkedIn/X automation and universal app control are not part of the current alpha.
- Is this production-ready software?
- No. Velyn is in private alpha: expect rough edges, changing features, and no uptime guarantee. Request access only if you are comfortable testing early software on your Mac.
- What does “private” mean here?
- Local-first AI on your Mac, approval before sensitive actions, and inspectable execution. It does not mean zero network use or a legal privacy guarantee—see the Privacy page for intent-level detail.
Access
Request early access to Velyn AI.
Early access requests are handled through BuildLayer, the product studio building Velyn. The form opens with Velyn alpha details prefilled.
Request early access via BuildLayer ↗