Free, open source, local AI for everyone.

icanhelp is a free, open-source Linux desktop AI assistant designed to be simple and user-friendly, running entirely on your own hardware. No GPU required, no cloud bill, no account needed.

~600 MB smallest model
0 GPUs required

Features

Runs on old laptops

No GPU? No problem. icanhelp runs on CPU with models as small as ~600 MB, making it perfect for older laptops, repurposed machines, or any hardware you already own. Just a few hundred MB on disk.

Easy on the planet

Fully offline and running on your own hardware, CPU or your computer's GPU, means no distant datacenters burning power for your questions.

Runs fully offline

Local models handle chat and vision. No account, no cloud lock-in, no internet needed for everyday use.

Model download manager

Pick and fetch models from a built-in list. Qwen 3.5 2B is recommended to start.

File & screenshot capture

Attach files or capture the screen so the assistant can see what you see.

Bash command execution

Let it run commands on your behalf, with confirmation prompts for anything risky.

Web search & vision

Search the web when needed and use a local vision model to understand images.

Addon models

Download specialized models for math computation. They load on-demand via the Addons panel.

Theme system

Light and dark themes, with a custom theme editor to match your desktop.

Buddy skins

Customize your AI avatar with different states — idle, talking, thinking, bash, and search. Import custom skin packs via ZIP.

Document creation

Generate .docx Word documents on the fly. Describe what you need and get a formatted file ready to share.

Chat history

Conversations are saved automatically. Browse past chats, pick up where you left off, or delete old threads.

Addons

Extend icanhelp with specialized models that load on-demand for specific tools. Download them from Settings → Addons in the app. Each model consumes additional RAM while active but stays unloaded when not in use.

LFM2 350M Math Math
Utility ~210 MB

Math computation model for evaluating expressions, solving equations, and arithmetic. Lets the main LLM offload computation to a specialized model.

Needs ~4 GB RAM

BLIP Image Captioning Vision
Utility ~380 MB

Vision model for describing images. Used by OCR and screenshot analysis.

Needs ~2 GB RAM

How addons work

1

Download addon models from Settings → Addons in the app.

2

Models stay idle on disk until a tool needs them — no RAM wasted.

3

When you use tools, the right model loads automatically.

Install

One command clones the repo, installs dependencies, generates icons, and registers icanhelp as a system app.

bash
$ curl -fsSL https://raw.githubusercontent.com/TheShovel/icanhelp/main/install.sh | bash
$ git clone https://github.com/TheShovel/icanhelp.git
$ cd icanhelp
$ npm install
$ npm start

Requirements

  • Node.js Node.js 20+
  • npm npm
  • git git
  • rsync rsync
  • curl curl
  • unzip unzip

Compatibility

Architecturex86_64, ARM64
Libcglibc, musl
DistrosArch-based, Ubuntu-based (best); Fedora, openSUSE, Debian
Desktoprequired
Not supportedAlpine, WSL headless, 32-bit ARM

First launch

The setup panel appears. Pick a model from the download list and you're ready.

Our stance on AI

We don't support AI-generated "art". We see AI as a practical tool, meant to simplify tedious tasks and make life easier. Not to replace the things we enjoy doing. We will never add features that let you generate images, music or videos. All art for this project is made by TheShovel and the icanhelp contributors, without any generative AI tools.