Skip to main content

Search events with AI agents

AI summary
Learn how to integrate Imply Lumi with AI agents that support MCP to query Lumi events using natural language from desktop apps, CLI, and code editors.

About AI summaries.

Imply Lumi integrates with AI agents that support the Model Context Protocol (MCP). Once you connect an agent to the Lumi MCP server, you can query Lumi events using natural language from your desktop app, CLI, browser, or code editor. The agent translates your prompts into Lumi queries, runs them, and returns results.

This topic explains how to connect each supported agent to Lumi. If your agent isn't listed, see Other MCP clients.

Prerequisites​

To query Lumi events with an AI agent, you need the following:

  • Access to Lumi with the Data manager role or higher. For information on roles and permissions, see Use roles to control access.
  • To authenticate with an IAM key, a Lumi IAM key. You can select an existing key or create one on the MCP integration page. For details, see Create an IAM key.
  • An installed AI agent that you're logged in to. See the agent's own documentation for installation details.

Choose an authentication method and AI agent​

You can authenticate your agent with Lumi in one of two ways:

  • OAuth: Your agent guides you through signing in to Lumi. The agent provisions an IAM key on your behalf. We recommend OAuth when you query Lumi yourself.
  • IAM key: You add an IAM key to your agent's configuration manually. The agent queries Lumi with the IAM key's permissions. Use an IAM key for automated workflows that run without user interaction, or for agents or plans that don't support OAuth. Most agents store the IAM key token in plain text in their configuration file. Don't share or commit the file.

Claude.ai supports OAuth only.

The instructions on this page use tabs for each authentication method. When you select a tab, all sections on the page switch to that method.

Supported agents​

Lumi supports the following AI agents. The table shows how you configure each one and where the configuration goes.

AgentConfiguration methodConfiguration location
Antigravity CLIJSON configuration file~/.gemini/config/mcp_config.json
Claude Codeclaude mcp add commandClaude Code CLI
Claude Desktop and Claude.aiCustom connector or desktop extension with an IAM keyClaude connector settings
CodexTOML configuration file~/.codex/config.toml
CursorJSON configuration file~/.cursor/mcp.json
VS CodeJSON configuration file.vscode/mcp.json in your workspace
Other MCP clientsJSON configuration fileYour agent's MCP configuration file

Get the configuration from Lumi​

Lumi generates the configuration for each agent. To get it:

  1. From the Lumi navigation menu, click Integrations > Model Context Protocol (MCP).
  2. Under Select an authentication method, select OAuth or IAM key.
  3. If you selected IAM key, click Select or create key and select or create the IAM key for your agent.
  4. Under Select an AI agent, select your agent.
  5. Follow the instructions for your agent in the Lumi UI or in the following sections.

MCP integration page

All agents except the Claude Desktop extension connect to the Lumi MCP server at https://lumi.imply.io/mcp/v1. This URL appears in each configuration on this page.

Connect your agent​

Follow the instructions for your agent. Select the OAuth or IAM key tab to match the authentication method you chose in Lumi.

Antigravity CLI​

Antigravity CLI reads MCP servers from ~/.gemini/config/mcp_config.json. It uses serverUrl rather than url for the server address.

To connect Antigravity CLI to Lumi:

  1. Add the imply-lumi server to ~/.gemini/config/mcp_config.json. If the file doesn't exist, create it with the following content:

    {
    "mcpServers": {
    "imply-lumi": {
    "serverUrl": "https://lumi.imply.io/mcp/v1"
    }
    }
    }
  2. Restart Antigravity CLI.

  3. Run /mcp to open the MCP manager.

  4. Select imply-lumi and sign in to Lumi when prompted.

For more information, see MCP in the Google Antigravity docs.

Claude Code​

By default, Claude Code adds MCP servers to the current project only. To make Lumi available in all your projects, include --scope user in the command.

To connect Claude Code to Lumi:

  1. Run the following command to add the Lumi MCP server:

    claude mcp add -t http --scope user imply-lumi https://lumi.imply.io/mcp/v1
  2. In Claude Code, run /mcp to open the MCP manager.

  3. Select imply-lumi, then select Authenticate.

  4. Sign in to Lumi when prompted.

For more information, see Connect Claude Code to tools via MCP.

Claude Desktop and Claude.ai ​

To connect Claude to Lumi:

How you add the Lumi connector depends on your Claude plan:

  • Team and Enterprise plans: An organization owner adds the custom connector for the organization. Members then find the Imply Lumi connector in their connector settings and connect it.
  • Free, Pro, and Max plans: You add the custom connector yourself.

To add and connect the Lumi connector:

  1. Add Lumi as a custom connector using the URL https://lumi.imply.io/mcp/v1. On Team and Enterprise plans, an organization owner does this step.
  2. In a conversation, enable the Imply Lumi connector.
  3. Sign in to Lumi when prompted.
  4. Before you query Lumi, enable Imply Lumi in the conversation.

If you can't add a custom connector on your plan, use an IAM key with the Claude Desktop extension instead.

For more information, see Get started with custom connectors using remote MCP.

Codex​

Codex shares the configuration file ~/.codex/config.toml across the Codex CLI, IDE extension, and desktop app.

To connect Codex to Lumi:

  1. Add the following to ~/.codex/config.toml:

    [mcp_servers.imply-lumi]
    url = "https://lumi.imply.io/mcp/v1"
  2. Run the following command to authenticate with the Lumi MCP server:

    codex mcp login imply-lumi
  3. Sign in to Lumi when prompted.

  4. Start a Codex session and run /mcp to confirm imply-lumi is connected.

For more information, see Model Context Protocol in the Codex docs.

Cursor​

To connect Cursor to Lumi:

  1. Add the following to ~/.cursor/mcp.json. Create the file if it doesn't exist.

    {
    "mcpServers": {
    "imply-lumi": {
    "url": "https://lumi.imply.io/mcp/v1"
    }
    }
    }
  2. Restart Cursor.

  3. Sign in to Lumi when prompted.

For more information, see Installing MCP servers in the Cursor docs.

VS Code​

You query Lumi from VS Code through GitHub Copilot Chat, so you need an active GitHub Copilot subscription.

To connect VS Code to Lumi:

  1. In your VS Code workspace, create a file named .vscode/mcp.json with the following content:

    {
    "servers": {
    "Imply Lumi": {
    "url": "https://lumi.imply.io/mcp/v1"
    }
    }
    }
  2. Click Start above the Imply Lumi entry in mcp.json.

  3. Sign in to Lumi when prompted.

  4. Confirm the server status changes to Running.

Open GitHub Copilot Chat and set the mode to Agent, so Copilot can call the Lumi MCP server.

For more information, see Add MCP servers in VS Code.

Other MCP clients​

Many agents accept the standard mcpServers JSON format. If your agent isn't listed on this page, you can add Lumi to your agent's MCP configuration file, as long as your agent supports remote HTTP MCP servers.

Check your agent's documentation for the file location and any differences in format. For example, some agents use serverUrl instead of url, or require "type": "http" in the server entry. If the file already lists other servers, add the imply-lumi entry under mcpServers.

To connect your agent to Lumi:

OAuth requires an agent that supports MCP authentication. If your agent doesn't, use the IAM key method instead.

  1. Add the following to your agent's MCP configuration file:

    {
    "mcpServers": {
    "imply-lumi": {
    "url": "https://lumi.imply.io/mcp/v1"
    }
    }
    }
  2. If your agent is running, restart it.

  3. Sign in to Lumi when prompted.

Query Lumi​

Once your agent is connected, ask questions about your Lumi events in natural language. For example:

  • "What are the top 10 sources this week?"
  • "Show me the top 10 most requested URLs in the past hour."
  • "What were the top 5 URI paths accessed in the past week?"

The agent translates your prompt into a Lumi query, runs it, and returns the results. Most agents also show the query they ran. If your agent asks for permission to use the Lumi tools exposed by the MCP server, allow it. If your agent doesn't use the Lumi tools, mention Lumi in your query, for example "Use Imply Lumi to find the top 5 URI paths this week."

For example, here's a sample query and response in Cursor:

Cursor example query

Learn more​

See the following topics and third-party documents for more information: