MCP Server

Laritor provides a Model Context Protocol (MCP) server that lets AI assistants retrieve Laritor data and help investigate and fix exceptions from your local project.

Getting started

Laritor uses OAuth to securely connect your AI assistant to your account. When you first connect, your AI assistant opens a Laritor authorization page. Sign in if needed, then authorize the assistant to access your Laritor account.

Connecting Laritor

If your account is in the US (NYC), run:

claude mcp add --transport http laritor https://nyc.laritor.com/mcp

If your account is in the EU (Amsterdam), run:

claude mcp add --transport http laritor https://ams.laritor.com/mcp

Then run /mcp in a Claude Code session to complete authentication.

Add the Laritor debugging policy below to CLAUDE.md in your project root.

If your account is in the US (NYC), run:

codex mcp add laritor --url https://nyc.laritor.com/mcp

If your account is in the EU (Amsterdam), run:

codex mcp add laritor --url https://ams.laritor.com/mcp

Add the Laritor debugging policy below to AGENTS.md in your project root.

Add the following to .cursor/mcp.json if your account is in the US (NYC):

{
  "mcpServers": {
    "laritor": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://nyc.laritor.com/mcp"]
    }
  }
}

Create .cursor/rules/laritor-debugging.mdc with alwaysApply: true frontmatter, then paste the Laritor debugging policy below into the file.

If your account is in the EU (Amsterdam), use this URL instead:

{
  "mcpServers": {
    "laritor": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://ams.laritor.com/mcp"]
    }
  }
}

In Devin, go to Settings > MCP Marketplace and select Add Your Own. Choose HTTP as the transport and OAuth as the authentication method.

Enter this server URL if your account is in the US (NYC):

https://nyc.laritor.com/mcp

Or enter this server URL if your account is in the EU (Amsterdam):

https://ams.laritor.com/mcp

Save the server, then use Test listing tools to connect and complete authorization.

In Settings & Library > Knowledge, create a knowledge item named Laritor debugging policy. Use a trigger such as “when debugging, diagnosing, investigating, or fixing runtime issues,” then paste the policy below as its content.

Add the following to .gemini/settings.json if your account is in the US (NYC):

{
  "mcpServers": {
    "laritor": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://nyc.laritor.com/mcp"]
    }
  }
}

Add the Laritor debugging policy below to GEMINI.md in your project root.

If your account is in the EU (Amsterdam), use this URL instead:

{
  "mcpServers": {
    "laritor": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://ams.laritor.com/mcp"]
    }
  }
}

Add the following to .junie/mcp/mcp.json if your account is in the US (NYC):

{
  "mcpServers": {
    "laritor": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://nyc.laritor.com/mcp"]
    }
  }
}

If your account is in the EU (Amsterdam), use this URL instead:

{
  "mcpServers": {
    "laritor": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://ams.laritor.com/mcp"]
    }
  }
}

Add the Laritor debugging policy below to .junie/AGENTS.md in your project root.

Run opencode mcp add, select remote when prompted, and enter the server URL for your account region:

US (NYC): https://nyc.laritor.com/mcp
EU (Amsterdam): https://ams.laritor.com/mcp

Then authenticate the server:

opencode mcp auth laritor

Add the Laritor debugging policy below to AGENTS.md in your project root.

Open the Command Palette and select Kiro: Open workspace MCP config (JSON). Add the following to .kiro/settings/mcp.json if your account is in the US (NYC):

{
  "mcpServers": {
    "laritor": {
      "url": "https://nyc.laritor.com/mcp"
    }
  }
}

Create .kiro/steering/laritor-debugging.md with inclusion: always frontmatter, then paste the Laritor debugging policy below into the file.

If your account is in the EU (Amsterdam), use this URL instead:

{
  "mcpServers": {
    "laritor": {
      "url": "https://ams.laritor.com/mcp"
    }
  }
}

Open the Command Palette (Cmd/Ctrl + Shift + P), select MCP: Add Server, then select a remote HTTP server. Enter the server URL for your account region:

US (NYC): https://nyc.laritor.com/mcp
EU (Amsterdam): https://ams.laritor.com/mcp

Add the Laritor debugging policy below to .github/copilot-instructions.md in your project root.

For other MCP-compatible tools, use the server URL for your account region:

US (NYC): https://nyc.laritor.com/mcp
EU (Amsterdam): https://ams.laritor.com/mcp

If your tool does not support remote MCP servers natively, use mcp-remote with the URL for your region:

npx -y mcp-remote https://nyc.laritor.com/mcp

# EU accounts
npx -y mcp-remote https://ams.laritor.com/mcp

Add the Laritor debugging policy below to the project instruction file used by your AI tool. If it supports the AGENTS.md convention, create or update AGENTS.md in the project root.

Laritor debugging policy

Add the following Markdown to the instruction file for your AI agent. It ensures the agent consults Laritor production telemetry early when investigating runtime problems.

## Laritor debugging policy

This repository is connected to Laritor through MCP.

Whenever the user asks to debug, diagnose, investigate, or fix any of the following:

- exceptions
- bugs
- runtime errors
- HTTP 500 errors
- crashes
- failed requests
- unexpected application behavior
- production issues

you MUST consult Laritor early in the investigation, even if the user does not explicitly mention Laritor.

Required workflow:

1. Read `.laritor.json` to determine the Laritor application mapped to this repository.
2. Call the Laritor exception search tool for recent relevant exceptions.
3. If a relevant exception exists, retrieve its detailed occurrence information.
4. Carefully inspect the Laritor execution timeline, including queries, logs, cache operations, outbound requests, jobs, mails, notifications, and other recorded events.
5. Treat the ordered event timeline as primary debugging evidence because it shows what the application executed before and during the failure.
6. Correlate Laritor production evidence with the local source code before proposing or implementing a fix.
7. Do not rely solely on local logs, git history, source inspection, or tests when Laritor production telemetry is available.

Do not wait for the user to say "use Laritor".

Skip Laritor only when:
- the user explicitly asks you not to use Laritor, or
- the task cannot involve application runtime behavior, such as formatting code, renaming a variable, or editing documentation.

Select your application

The first time you use Laritor MCP in a local project, your AI assistant prompts you to select the Laritor application that belongs to the project. Once selected, it creates a .laritor.json file in the project root.

Laritor uses this file for later prompts, so you do not need to select the application each time.

Example workflows

Investigating an exception

Ask your AI assistant:

“Help me understand why the “trying to access array offset on null” exception is occurring.”

Your assistant can retrieve the relevant Laritor data and help you analyze the exception from your local project.

Triaging recent exceptions

Ask your AI assistant:

“What are the most recent exceptions in Laritor in the past 2 days?”

Resolving an exception

Ask your AI assistant:

“Fix the most recent exception in Laritor.”

Your assistant can use the connected Laritor data to investigate the problem and help you implement a fix.

Security

Authentication: Laritor MCP uses OAuth to securely connect to your account.
Authorization: The MCP server can access only the applications and data for teams you belong to.
Read-only access: The MCP server has read-only access to your Laritor account.