Connect Lightup to MCP-compatible AI agents.

This repository is the entry point for using Lightup inside agentic workflows. Client-specific setup lives in dedicated folders so the structure stays stable as support expands across interfaces.

Use the repo-level setup wrapper and pass the client you want to configure:

curl -sL https://raw.githubusercontent.com/lightup-data/lightup/main/setup.sh \
  | bash -s -- claude

curl -sL https://raw.githubusercontent.com/lightup-data/lightup/main/setup.sh \
  | bash -s -- gemini

The setup flow automatically looks for lightup-api-credential*.json in common locations such as ~/Downloads. If multiple files are found it picks the most recently modified one and warns you. If no file is found, the script will ask whether you have a Lightup account — if not, it walks you through signing up for a 30-day free trial at my.lightup.ai.

To use a specific credential file, pass the path explicitly:

curl -sL https://raw.githubusercontent.com/lightup-data/lightup/main/setup.sh \
  | bash -s -- claude /path/to/lightup-api-credential.json

Client-specific installation and usage details live in the dedicated guides below.

If you use Claude Code, you can install Lightup as a Claude Code plugin instead of running the setup script. The plugin auto-discovers your lightup-api-credential.json, connects automatically, and adds built-in slash commands and a specialized data quality agent — no credential prompts.

claude plugin marketplace add lightup-data/lightup
claude plugin install lightup-ai@lightup

The plugin finds your credential file in ~/Downloads, ~/Desktop, or ~ automatically. See lightup-ai-plugin/README.md for full details.

If you use Gemini CLI, you can install Lightup as a Gemini CLI extension instead of running the setup script. The extension auto-discovers your lightup-api-credential.json, connects automatically, and adds a specialized data quality agent — no credential prompts.

gemini extensions install https://github.com/lightup-data/lightup

The extension finds your credential file in ~/Downloads, ~/Desktop, or ~ automatically. See lightup-gemini-plugin/README.md for full details.

• Claude Code
• Gemini CLI
• _{Codex CLI (coming soon)}

Lightup helps teams bring trusted data quality context into the tools they already use to investigate issues, debug pipelines, and answer operational questions. This repo packages client-specific setup instructions and scripts in one place, starting with Claude Code and designed to extend cleanly to additional AI agent clients over time.

• Claude Code: available now
• Gemini CLI: available now
• Codex CLI: coming soon

Each client guide is responsible for its own installation and connection flow. That keeps the top-level repository product-oriented, while letting each integration evolve independently.

Pick the guide for your preferred AI client to get started.

Lightup Agentic exposes your Lightup data quality platform as an MCP (Model Context Protocol) server — letting any MCP-compatible AI agent read metrics, monitor data quality, diagnose incidents, and create new monitors in plain English, without leaving your AI tool.

┌─────────────────────┐        MCP (SSE)        ┌─────────────────────┐
│   AI Agent Client   │ ◄─────────────────────► │  Lightup MCP Server │
│                     │                          │                     │
│  • Claude Code      │                          │  Exposes 41 tools   │
│  • Gemini CLI       │                          │  for metrics,       │
│  • Codex CLI        │                          │  monitors,          │
│                     │                          │  incidents, docs    │
│                     │                          └──────────┬──────────┘
└─────────────────────┘                                     │
                                                       REST API
                                                            │
                                               ┌────────────▼───────────┐
                                               │   Lightup Platform     │
                                               │   (your instance)      │
                                               └────────────────────────┘

The Lightup MCP server is stateless. Each connection carries its own credentials in the URL — no credentials are stored server-side.

Your AI client connects with:
  /sse?host=https://app.acme.lightup.ai&refresh_token=eyJ...
         │                                      │
         ▼                                      ▼
  Your Lightup instance URL         Your JWT refresh token
  (which Lightup instance to use)   (your identity)

Every session is fully isolated — the server only sees and accesses your Lightup instance using your credentials. No data is shared across sessions.

Security: Credentials travel over HTTPS only and are never logged or stored on the MCP server.