Artificial intelligence news.
Straight from the original announcements.

webAI launches Mac app for private on-device AI without per-token charges

November 5, 2026, 23:47By the AI News Bites Editorial TeamSource: PR Newswire
webAI launches Mac app for private on-device AI without per-token charges
Image: PR Newswire

webAI has released its Mac app to the public, allowing users to run the company’s small reasoning models on their own computers without sending files to the cloud or paying per token. The app also lets users build teams of AI experts that work across Slack, DocuSign, Hugging Face and Apple apps.

Key points
  • webAI released its app publicly for Mac on Nov. 5, 2026.
  • The app runs webAI’s small reasoning models on the user’s computer.
  • Users can work with private files without sending them to the cloud.
  • The app has no per-token charges and no token limit for repeated revisions.
  • Personas can connect to Slack, DocuSign, Hugging Face and Apple apps.

AI processing stays on the Mac

The webAI app runs its models on the user’s Mac, which the company says allows people to work with private documents without sending them to online services. Users can revise a sensitive client proposal or an unfinished business plan repeatedly without reaching a token limit.

webAI positions the app as an alternative to cloud-based AI services, where usage can be metered by tokens or limited by a plan, and documents are processed on another company’s servers. The company says its local approach gives users control over what they share while removing per-token charges for work handled by the app.

“With the webAI app, we're making cloud AI optional,” said David Stout, CEO and co-founder of webAI. “Why should you have to send a private document to an online service just to get help with it? Your Mac can do the work.”

Users can assemble teams of AI Personas

The app’s main building blocks are called Personas. Users create each Persona with a role, knowledge and tools of their choosing, then control which information it can access and which actions it can take.

Several Personas can work on the same task. In webAI’s example, one Persona could pull client requirements from project notes, another could draft a proposal and a third could check the draft for missing information.

Connectors link the Personas with services and apps including Slack, DocuSign, Hugging Face and Apple apps. A Persona could retrieve a client’s latest decisions from Slack, identify unsigned contracts in DocuSign and prepare follow-up messages for the user’s approval. According to webAI, this can reduce the need to search for context and copy details between applications.

Available now for Mac

The webAI app is available now for Mac through webAI.com. The release marks the first time webAI has made the app available to the public.

The launch gives Mac users a way to apply AI to private work while keeping files on their own computers, according to the company. It also combines local model execution with controlled access to workplace tools, allowing users to decide what each Persona can see and do. webAI builds technology intended to let people create, run and control AI on their own devices.

Related news

Databricks makes Genie in Slack generally available
Products and toolsNovember 12, 2026, 16:00

Databricks makes Genie in Slack generally available

Databricks has made its Genie app for Slack generally available, allowing users to ask governed questions in channels, threads and direct messages without leaving Slack.…

BotPsy brings a 0.8B on-device agent to local AI workflows
ModelsNovember 11, 2026, 06:01

BotPsy brings a 0.8B on-device agent to local AI workflows

Tether AI Research has introduced BotPsy, a 0.8B-parameter language model designed for function calling and agentic workflows on devices. Built on Qwen3.5-0.8B, the…

Hugging Face marks AnyLanguageModel 1.0 and a year of local Apple models
For developersOctober 20, 2026

Hugging Face marks AnyLanguageModel 1.0 and a year of local Apple models

Hugging Face has published a blog post titled "AnyLanguageModel 1.0: One Year of Local Models on Apple Platforms", dated 2026-10-20. The text available from the page…