Explicit tools
The planned runtime gives each workflow a defined tool set rather than assuming unrestricted access to your machine.
LOCAL AI Agents · Coming soon
Useful agents.
Visible boundaries.
A local agent runtime for workflows you can inspect and permissions you can understand. LOCAL AI Agents is coming soon.
Planned experience
Give local models a workflow.
A local agent runtime designed for explicit tools, clear permissions and work you can inspect.
The workflow above is a simulated illustration. It has no access to your files, tools or operating system.
LOCAL AI AgentsSIMULATED UI
Workflow inspectorExample run
Read project notesPermission requested
Draft a summaryLocal inference
Write to workspaceApproval required
Input local model
Tool request your approval
Output review
Tool request your approval
Output review
Planned workflow · all steps simulated
Product direction
Built around your control.
These are design intentions for the planned product. Availability and final scope will be confirmed at launch.
Human review
Actions that modify files or reach beyond the workspace should require deliberate approval. Exact permission behaviour will be specified before release.
Inspectable workflows
See the relationship between a prompt, tool request and result. Local execution should still be understandable to the person operating it.
Local does not mean harmless
Model output is untrusted input. Plan for prompt injection, sensitive files and unintended tool use before connecting an agent to real work.
Before you get started.
There are no installers, downloads or active endpoints in this preview. Compatibility, supported platforms and release requirements will be published with early access.
- Review model licences for your intended use.
- Plan memory, storage and operational responsibilities.
- Keep sensitive data out of unreviewed workflows.
Your AI. Your hardware.
Your next move.
Help shape LOCAL AI. Tell us what you want to run.