OpenViSTAI · Agents · Blockchain · Crypto
Menu
Workflow Agents

Klee

Klee is a native macOS AI assistant that runs supported language and vision models locally on Apple Silicon using MLX, with chat and built-in tools for file operations, web search, webpage retrieval and shell commands.

Independent OpenViST project profile, based on linked sources.

Klee — OpenViST project profile in Workflow Agents.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Klee when you have an Apple Silicon Mac meeting its macOS and memory requirements and want local model chat with desktop tools. Configure a Jina API key when web search is part of the task.

Who it is for

  • Mac users who want to chat with locally run language models.
  • People who want an assistant to read, create, list or delete local files through tool calls.
  • Users seeking web search and webpage retrieval from a desktop assistant, with a Jina API key for search.
  • Users who want to ask questions about images using supported vision-language models.

Documented uses

  • Ask questions in chat using a model downloaded to the Mac.
  • Read, create, overwrite, list or delete files through built-in file tools.
  • Search the web with a configured Jina API key and retrieve webpage content.
  • Run local shell commands through the shell tool.
  • Attach images and ask questions when using a supported vision-language model.
Before you choose

Requirements & limits

The app is distributed as a signed macOS app through GitHub Releases, not the App Store. The README states that local chat requires no account or API key; web search requires a Jina API key. The documented minimum is macOS 15 or later, Apple Silicon M1 or later, and 16 GB RAM.

  • Hardware requirements are specific: macOS 15+, Apple Silicon M1 or later, and at least 16 GB RAM.
  • The README’s model recommendations scale with memory: 16 GB for several smaller options, 32 GB for listed 27B and 35B models, 64 GB for 27B–32B options, and 96 GB or more for the listed 122B model.
  • Web search depends on a Jina API key; local chat does not.
  • Vision support is limited to supported vision-language models, identified in the model list.
  • The shell tool has a documented 30-second timeout.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Klee and start a local model chat

Install the macOS app, download a recommended model and begin a chat.

Before you start

  • macOS 15 or later
  • Apple Silicon Mac (M1 or later) with at least 16 GB RAM

Steps

  1. Download the latest DMG from the GitHub Releases page.

  2. Drag Klee into the Applications folder and open it. If Gatekeeper displays a warning, use System Settings > Privacy & Security > Open Anyway.

  3. Choose a model recommended for the detected system memory and download it.

  4. Select the downloaded model and type a message to start chatting.

Expected outcome

A chat with a downloaded model running locally through MLX on the Mac.

Guide limitations

  • The available model choices depend on system memory; larger listed models require more RAM.
  • The documented installation is for macOS on Apple Silicon.
Source-based practical guide

Configure Klee web search

Enable web search through Jina AI and submit a search request from Klee.

Before you start

  • Klee installed and open
  • A Jina AI API key

Steps

  1. Get an API key from Jina AI.

  2. Open Klee’s sidebar using its toggle button.

  3. Enable Web Search and enter the API key.

  4. Ask Klee to search for a topic.

Expected outcome

Klee can use its web-search tool with the configured Jina API key.

Guide limitations

  • Web search requires a Jina API key; ordinary local chat does not require an API key.
Evidence and transparency

Sources & review

Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
signerlabs/Klee · Official homepage title · retained in its source language