AI NewsOpen sourceAnnouncement

Linger puts a local personal agent on your own PC

Linger is a local personal agent that opens real apps on a Mac, Windows PC or Linux desktop. The project ranked 5th on Trendshift's daily list and has 1,234 stars.

AI News

Editorial2 min read

LinkedInX
GitHub social card for the simplexai-labs/linger repository

Image: GitHub

Why it mattersMost consumer AI runs in somebody else's cloud with your account. An agent that opens your own browser and writes to your own files is a real option again.

A personal AI agent is supposed to do your work while you do something else. Most of today's agents live in a vendor's cloud. They sign in to that vendor's accounts. Linger, an open-source project from simplexai-labs, is a try at the other shape. It runs on one PC, opens the apps already installed on it, and keeps the data on the same disk.

The repository ranked fifth on Trendshift's daily list today, and holds 1,234 stars at the time of writing. The README calls it "an Open-Source Personal Agent That Runs Fully on Your PC", with "a Claude-class brain running locally". The project's stated goal is to put powerful local AI on consumer machines, through what it calls model-harness co-design.

What it actually does

Linger controls the desktop the user already has. The README states: "Your apps, your Chrome, already signed in." Agents open those apps on Mac, Windows or Linux. A web agent in the cloud clicks through fresh browser sessions. Linger inherits the user's logged-in state instead. It uses the same Google account the user already uses, with no password handover.

Three capabilities read as the point of the project. The first is proactive work. The README says "Another Me watches your inbox, calendar and Slack, follows up and moves work forward". The second is shared memory across agents. "Tell one agent, and every agent knows", through wiki pages updated nightly. The third is a team pattern inside one PC. "One plan for every agent" means a user can delegate to several at once.

The local trick

The reason this works on a laptop is an inference engine the project calls PolyStrata. PolyStrata "keeps a model's busiest experts on your GPU and the rest in RAM or on SSD", the README writes. The headline number is specific. A 125B-parameter model, Qwen3.8-Flash-Next, "runs on a 12 GB card with 64 GB of RAM", by the project's own measurement.

The caveats are the project's own too. That is simplexai-labs's own statement about its own code, and the number is not independent. Any team building on top of Linger should run the model on its own hardware before planning around the figure. A 125B model on a consumer card is still slow in absolute terms, even when it fits.

Who this is for

Linger is early. 1,234 stars in a month is real traction but not a user base. A desktop agent that opens Chrome depends on each app behaving the same way next week. The reader to care about this today is not a startup rolling out an AI product. It is a knowledge worker who wants an inbox agent without handing an outside company their inbox. It is also a developer weighing the local option for a client who worries about data leaving the laptop.

Source

GitHub, simplexai-labs/linger: Linger is An Open-Source Personal Agent That Runs Fully on Your PC.

SourceGitHub

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project