The discovery engine for open-weight and open-source AI.

From models to apps: three waves of open AI

Open-weight and open-source AI has moved in three waves. Each one solved a problem and exposed the next.

1. Wave 1

   Hugging Face
2. Wave 2

   Ollama & LM Studio
3. Wave 3

   Apps

1. Wave 1

   Models

   Made AI downloadable.

   Open-weight and open-source models (Llama, Qwen, Mistral, Gemma, DeepSeek and hundreds more) went from research artifacts to files anyone can download. Hugging Face is where they live.

   The catch: A model file cannot answer a question by itself.
2. Wave 2

   Runtimes

   Made models runnable.

   Ollama, LM Studio and llama.cpp turned a download into a running model on your own machine. Type a prompt, get an answer.

   The catch: Usage stays narrow. For most people it is a chat window, and many stop there.
3. Wave 3

   Applications

   Makes local AI useful.

   Apps built on top: chat with your documents, code in your editor, run agents, transcribe and speak, generate images and video, automate your home. They work with the model directly or through Ollama, LM Studio and other interfaces.

   The catch: There are hundreds of them, scattered across GitHub READMEs, Discord threads and changelogs, mostly in English and quickly outdated.

   This is where PromptQuorum works

Models are the engine. Runtimes start the engine. Applications are what people actually drive. PromptQuorum is the map, the manual and the test drive for that third wave.

Why local: what we believe

• Local models are open. You can download the weights and run them on your own hardware. No one has to approve your prompt, and no one sits between you and the model.

• Cost stays in your control. Once a model runs on your machine, there is no meter, no per-token bill and no price change you did not agree to. The hardware is the cost, and you decide how much to spend on it.

• It can’t be taken away. A cloud model can be changed, restricted, repriced or shut down overnight. A model you have downloaded keeps working the same way as long as you keep the file.

• Your data stays with you. When both the app and the model run locally, your prompts and documents never leave your machine.

• Independence is the point. Local AI is more than a cheaper alternative. It is a way to depend on no one else’s terms, servers or decisions.

Independence from control is the theme of everything on this site. It is why the directory is neutral, why every entry links to the official source and why we publish in nine languages.

We don’t believe cloud AI is bad. We believe you should have a real choice, and that choosing local should be as easy as choosing cloud.

One ecosystem, four parts that work together

PromptQuorum is not just an app, and not just a directory. It is four parts built around the same goal, so you can go from “what should I run?” to a working setup without leaving the site.

Start with a goal, not a product name

You should not need to know that “llama.cpp” or “Open WebUI” exist before you can find them. Start with what you want to do:

• I want a private ChatGPT alternative on my laptop.
• I have 16 GB of RAM. What can I realistically run?
• I want to chat with my own PDFs offline.
• I need a local coding assistant for VS Code.
• I want an offline voice assistant for my smart home.

Use PromptQuorum from your AI

PromptQuorum is the first AI app directory to publish a public, no-authentication MCP endpoint. Connect it once, and any MCP-compatible assistant — Claude, ChatGPT, Cursor and others — can query PromptQuorum’s app database, license data and articles directly inside your own conversation, no copy-pasting between tabs.

That data becomes an input your assistant can reason over, not just read. It combines what PromptQuorum returns with its own knowledge, the other files in your conversation and its own web research, in the same reply. In practice, that lets you:

For example: ask "which local AI apps run on 16 GB of RAM and can chat with PDFs?" and your assistant queries PromptQuorum's directory and knowledge base directly, then names specific apps with their license and hardware fit — instead of guessing from training data that may be out of date.

• Compare apps PromptQuorum has reviewed without leaving your chat, then ask follow-up questions about them
• Check a tool’s license against your use case, cited straight from PromptQuorum’s data
• Cross-reference PromptQuorum’s hardware requirements against your own specs or a document you’ve uploaded
• Fold PromptQuorum’s directory data into broader research your assistant is already doing on the web

Connect PromptQuorum to your assistant once, and it can look things up for you.

MCP endpoint:

https://www.promptquorum.com/api/mcp (public, read-only, no authentication)

Discovery manifest:

https://www.promptquorum.com/mcp.json

{
  "mcpServers": {
    "promptquorum": {
      "url": "https://www.promptquorum.com/api/mcp"
    }
  }
}

Available tools: search_promptquorum, get_article, list_clusters, get_app_details, explain_license, search_apps, list_categories