Skip to content

Quick Start

Manuel Luján edited this page Nov 6, 2026 · 1 revision

Quick start

Four ways in, from the quickest to the most involved. The first two need no Fabric access at all.

1. Demo image (no install, no credentials)

docker run --rm -p 8765:8765 ghcr.io/largolujan/lineageiq:latest

Open http://localhost:8765. The "Demo data" menu in the top-right corner switches between English and Spanish and between light and dark themes.

The image holds only the public code and the synthetic Contoso tenant, serves reads only, and has the remote /mcp endpoint switched off. Every release is smoke-tested by the demo-image workflow before it is published. Tags: latest, and one per release (for example 0.2.0 and 0.2).

To use the MCP tools from Claude Desktop, Claude Code or any other MCP client, register the stdio server inside the image:

{
  "mcpServers": {
    "lineageiq-demo": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "ghcr.io/largolujan/lineageiq:latest", "python", "backend/scripts/mcp_server.py"]
    }
  }
}

Questions to try:

  • "What feeds the RPT_Sales_Dashboard report?"
  • "What depends on the sales_summary table?"
  • "Which items fail most often?"

2. From source, with demo data

Requires Python 3.12 and Git. The viewer and the demo generator use only the standard library.

git clone https://github.com/LargoLujan/lineageiq.git
cd lineageiq
python backend/scripts/generate_demo_data.py
LINEAGE_DEMO_MODE=1 python backend/scripts/lineage_web_server.py --bind 127.0.0.1 --directory frontend/semantic-lineage

On Windows PowerShell, set the variable first: $env:LINEAGE_DEMO_MODE = "1", then run the server command without the prefix. Open http://127.0.0.1:8765.

For the MCP servers, also run pip install -r requirements-mcp.txt. Details: Quick Start With Demo Data in the README.

3. Your own tenant

Never use LINEAGE_DEMO_MODE here: it makes every read public.

  1. pip install -r requirements-web.txt.

  2. Copy config/admin_settings.example.json to config/admin_settings.json and set tenant_id and the admin allowlist. An Entra app registration is optional: without one, sign-in uses the Azure CLI's public client ID.

  3. Run discovery, either:

    • locally: az login --use-device-code --tenant <tenant-id> --allow-no-subscriptions, then pwsh -NoLogo -File backend/scripts/99_refresh_lineage_project.ps1 -Mode full-read-only (needs PowerShell and the Azure CLI), or
    • inside Fabric: import the notebooks from fabric_notebooks/ and chain them in a Data Pipeline, as described in Fabric-native discovery setup.
  4. Start the server and sign in with your Microsoft account:

    python backend/scripts/lineage_web_server.py --directory frontend/semantic-lineage

Read next: Pointing It at Your Own Tenant in the README, the admin panel and authentication guide and the tooling prerequisites.

4. Connect an AI agent to a running server

You want Use Credential
An agent on the same machine as the data backend/scripts/mcp_server.py (stdio) None for reads
An agent on your machine, data on a server backend/scripts/mcp_server_remote.py --base-url <server URL> (stdio) api_key argument on each tool
A hosted AI product that can only reach a URL https://<your-host>/mcp (Streamable HTTP) Authorization: Bearer <api-key>

API keys are created by an admin in the admin panel's "API keys" tab. The remote MCP setup guide has steps for Claude Code, Claude Desktop, ChatGPT, Cursor, Windsurf and GitHub Copilot, plus a curl check. The tools themselves are listed in the MCP tools reference.

Next

Clone this wiki locally