docs

This is the official documentation repository for Liquid AI. It contains comprehensive guides, API references, and tutorials for building with our open-weight LFMs on laptops, mobile, and edge devices. The documentation is hosted at docs.liquid.ai.

Table of Contents

• Connect Liquid Docs to AI Coding Tools
  • What is MCP?
  • Connect to Claude Code
  • Connect to Cursor
• Contributing to the Documentation
  • Prerequisites
  • Install the Mintlify CLI
  • Run the Documentation Locally
  • Making Changes
  • Submitting Changes
  • Link Check
  • Link Snapshot
• License

---

Connect Liquid Docs to AI Coding Tools

What is MCP?

The Model Context Protocol (MCP) is an open standard that gives AI applications a standardized way to connect to external data sources and tools. By connecting your AI coding tool to Liquid docs via MCP, you’re giving it live, queryable access to the complete documentation: not a snapshot, not a cached file, but a real-time search against your published content. Every time Liquid docs update, the MCP server reflects those changes automatically.

Liquid docs hosts an MCP server at:

https://docs.liquid.ai/mcp

---

Connect to Claude Code

Step 1: Add the MCP server

Run the following command in your terminal:

claude mcp add --transport http liquid-docs https://docs.liquid.ai/mcp

This will automatically configure Claude Code to connect to the Liquid docs MCP server.

Step 2: Verify the connection

First, confirm the MCP server was added by listing all configured servers:

claude mcp list

You should see liquid-docs in the list of MCP servers.

Then, in a Claude Code session, ask a question about Liquid AI:

How do I run an LFM model on Android with llama.cpp?

Claude Code will query your MCP server and return an answer grounded in Liquid docs.

---

Connect to Cursor

Option A: One-click install from Liquid docs (fastest)

1. Open any page on the Liquid docs site.
2. Click the contextual AI menu icon (✨) in the top-right corner of the page.
3. Select Connect to Cursor.
4. Cursor will open and prompt you to confirm. Click Allow.

Done - no manual configuration needed.

Option B: Manual setup via the Command Palette

Step 1: Open the Command Palette

Press Cmd+Shift+P on macOS or Ctrl+Shift+P on Windows.

Step 2: Search for MCP Settings

Type Open MCP settings and select it.

Step 3: Add a custom MCP server

Click Add custom MCP. This opens the mcp.json file.

Step 4: Add the configuration

Paste the following:

{
  "mcpServers": {
    "liquid-docs": {
      "url": "https://docs.liquid.ai/mcp"
    }
  }
}

Save the file and restart Cursor.

Step 5: Verify the connection

Open a Cursor chat and ask something about Liquid AI. Cursor will query your MCP server and use Liquid docs as context when generating responses.

---

Contributing to the Documentation

We welcome contributions to improve the Liquid AI documentation! Here’s how to get started:

Prerequisites

You’ll need Node.js installed on your machine.

Install the Mintlify CLI

Install Mintlify locally using npm:

npm i -g mintlify

Run the Documentation Locally

Navigate to the docs directory and start the development server:

cd docs
npm i
mintlify dev

The documentation will be available at http://localhost:3000 (or http://localhost:3001 if port 3000 is already in use).

Making Changes

1. Edit the .mdx or .md files in the docs directory
2. The dev server will automatically reload to reflect your changes
3. Follow the guidelines in CLAUDE.md for MDX syntax and styling conventions
4. Test your changes thoroughly before submitting a pull request

Submitting Changes

1. Create a new branch for your changes
2. Make your edits following our documentation standards
3. Test locally using mintlify dev
4. Submit a pull request with a clear description of your changes

For more details on Mintlify setup and configuration, visit the official Mintlify documentation.

Liquid API Reference

The API and API reference tabs document the hosted Liquid API served by Liquid Console. The endpoint pages under API reference are generated by Mintlify from api/openapi.json. Don’t edit that file by hand: it is exported from the endpoint registry in the console repository (apps/web/lib/api/docs.ts). After an API change lands there, regenerate it from the console’s apps/web directory and open a pull request here with the result:

bun run openapi:export <path to this repo>/api/openapi.json

The overview, quickstart, models page, and guides under api/ are written by hand.

Link Check

The check-docs.yaml workflow has a check-link job that examines markdown links. Customize the config in link-check.json. If a link cannot be accessed (e.g. GitHub private repo), add the URL pattern to the ignorePatterns array.

Link Snapshot

link-snapshot.yaml records every URL the docs site has served. The check-link-snapshot.yaml workflow fails any PR that would make a recorded URL stop resolving. Run npm install once to install the pre-commit hook that keeps the snapshot in sync.

How it works:

• link-snapshot.yaml has two sections: active: (URLs the site is currently committed to keep serving) and deleted: (URLs intentionally retired).
• The pre-commit hook regenerates the file whenever docs.json or any .mdx/.md file is staged.
• The active: section is append-only — new URLs are added when pages are created, but existing entries are never silently dropped, even after you delete the underlying page.
• CI re-checks every URL in active: on each PR. A URL is satisfied if it’s still in the navigation, still on disk, or reachable via a redirect chain.
• The only way to retire a URL is to move it from active: to deleted: by hand. This forces every removal to be a deliberate choice — redirect, deprecate-but-keep, or formally retire.

See CLAUDE.md for the remediation paths and the full resolution algorithm.

---

License

This documentation is licensed under Attribution-ShareAlike 4.0 International.