Linger Get Linger

Your agent. On your computer.

Find what you need. Remember what matters. Move work forward with an open-source personal agent on your own computer.

Get Linger Explore the product

Meet
Another Me.

Your personal agent connects the dots across your files, mail and calendar. It remembers what matters and prepares your next step—from a client follow-up to Monday’s plan.

Watch the overview
Another Me
“I found who can help with this project. Your email draft is ready.”
Project follow-upFor your review

Linger overview

Video description

Linger runs a personal agent on your computer. The overview shows how it finds project information across apps, remembers a birthday and a gift preference, prepares a weekly summary, and coordinates a team of agents.

One project.
Many sources.

Bring the latest email, Slack thread and meeting notes into one brief. Linger follows the project across the files and apps you connect.

Under the hood

Powerful inside.
Personal outside.

Connected context. Lasting memory. Local intelligence.

Connected context

Context connected.
Next step ready.

  • Files
  • Mail
  • Calendar
  • Slack
  • Docs
Linger

Example

Contract review pending.

Find the right person.

Shared wiki

Agent memory

Your apps. One clear next step.

How it works

Linger looks across the sources you authorize to connect the people, conversations and documents a project needs. Changing details are checked against their latest sources before it presents a next step.

Your personal wiki is a shared reference that all your agents can search. Each agent also keeps a separate memory of its own conversations, preferences and lessons.

Local mode requires a configured local model. Inference can run on your computer; connected apps still communicate with their services.

Architecture Wiki & memory

Local inference

A bigger model.
A smarter fit.

Big models, shared across your GPU, RAM and SSD.

125Bparameter model
12GBGPU memory

With 64 GB of system memory.

The busiest parts stay in fast graphics memory.

Select a layer

Technical details

Qwen3.8-Flash-Next · Q2_0. This 125B mixture-of-experts model uses a few specialists at a time. PolyStrata shares the work across GPU and system memory, with model files and lookup tables on disk. The diagram is illustrative, not a memory allocation chart.

93.0output tokens / secondProject-reported benchmark

RTX 5070 · 12 GB VRAM + 64 GB RAM
Qwen3.8-Flash-Next · Q2_0 · 4K prompt

See the benchmark

Test conditions

Output speed from the project’s published test: Ryzen 5 7600, 64 GB DDR5-5200, Windows, Strata 0.1.26, a 4K-token prompt and 256 generated tokens, with multi-token prediction enabled. Tokens are small pieces of text. This is a measured configuration, not a speed guarantee; performance varies with hardware and workload.

Explore the inference engine Meet the model

Get Linger.

A little less to keep in your head. A little more ready when you need it. Bring Linger to your computer.

macOSApple silicon · DMG Windowsx64 · EXE Linuxx64 · TAR.GZ
All releases Docker setup

Common questions.

Browse the docs
What can I ask Linger to do?

Start with a real task: “Who can help me move this project forward?”, “Recap this week’s open tasks for Monday”, or “Help me prepare the product launch materials.” Linger can find the context, prepare a draft and bring the next step back to you.

What stays on my computer?

With a local model and a local server, inference and memory stay on your machine. You choose which files and apps Linger can use. Connected apps still communicate with their own services.

Which computers can run Linger?

The desktop app runs on Apple silicon Macs, 64-bit Windows and 64-bit Linux with glibc 2.35 or later. Intel Mac users can use Docker Compose. Local model support depends on your hardware and available memory.

How does a larger model fit on a PC?

For supported mixture-of-experts models, PolyStrata keeps frequently used parts of the model in GPU memory and places others in system memory or on disk. Read about local inference.

Can several agents work on one task?

Yes. Linger can bring together a team for a single brief. A research agent, a writer and a visual agent can follow a shared plan and prepare a draft for you to review.

Where can I explore the source?

The code, release notes and contribution guide are in the Linger repository. The project publishes its terms in the Linger Open Source License.