Skip to main content

​
Overview

TierZero provides a remote Model Context Protocol (MCP) server that you can connect to any MCP-compliant client. This enables AI assistants like Claude Code or Cursor to access your production context directly, querying logs, metrics, traces, and other telemetry data from your connected integrations to help with debugging, incident investigation, and operational tasks.

​
Prerequisites

Before setting up the MCP server, you need to generate a Personal Access Token (PAT) from TierZero:
  1. Navigate to https://auth.tierzero.ai/account/api_keys/
  2. Create a new personal API key
  3. Add the key to your shell
export TIERZERO_PAT_TOKEN="<new-api-key>"
  1. Save this token to TIERZERO_PAT_TOKEN in ~/.zshrc or your preferred shell.
Keep your PAT secure! This token grants access to your production telemetry data. Never commit it to version control or share it publicly.

​
Setup Instructions

The easiest way to use TierZero with Claude Code is through our official plugin:
  1. Add the TierZero marketplace:
/plugin marketplace add TierZeroAI/tierzero-claude-code
  1. Install the plugin:
/plugin install tierzero@tierzero
Once installed, use the /tierzero command to investigate production issues: 
/tierzero What errors occurred in payment-service in the last hour?
This runs in the background so you can continue working while TierZero queries your telemetry. Use /tasks to check progress.

​
Project-Wide Installation

To enable the TierZero plugin for your entire team, add this to your project’s .claude/settings.json: 
{
  "extraKnownMarketplaces": {
    "tierzero": {
      "source": {
        "source": "github",
        "repo": "TierZeroAI/tierzero-claude-code"
      }
    }
  },
  "enabledPlugins": {
    "tierzero@tierzero": true
  }
}
Commit this file to your repository. Anyone who clones the project will have the TierZero plugin available automatically.

​
Alternative: Manual MCP Configuration

If you prefer manual configuration, you can add the MCP server directly: 
claude mcp add tierzero-server https://api.tierzero.ai/mcp/ \
  --transport http \
  --header "Authorization: Bearer \${TIERZERO_PAT_TOKEN}"

​
Setup for Cursor

For Cursor, you’ll need to manually configure the MCP server in your settings:
  1. Open Cursor Settings (Cursor → Settings → Cursor Settings)
  2. Navigate to Tools & MCP
  3. Add a new MCP server with the following configuration:
{
  "mcpServers": {
    "tierzero-server": {
      "type": "http",
      "url": "https://api.tierzero.ai/mcp/",
      "headers": {
        "Authorization": "Bearer ${env:TIERZERO_PAT_TOKEN}"
      }
    }
  }
}
  1. Restart Cursor to apply the changes

​
How to use

If you’re using Claude Code with the TierZero plugin installed, Claude Code will automatically load the tierzero skill when your task would benefit from production context. In practice, that means Claude Code can proactively query TierZero in the background while it continues working on your code. Claude Code automatically loading the TierZero skill and running a TierZero production context query in the background You can also explicitly start an investigation with: 
/tierzero What errors occurred in payment-service in the last hour?
Under the hood, the plugin calls the tierzero_ask MCP tool. Use it when you want TierZero to investigate and synthesize production context for you. It can:
  • Query logs, metrics, traces, and spans from connected integrations such as Datadog, New Relic, CloudWatch, Grafana Cloud, and Sentry
  • Inspect GitHub repositories, CI/CD systems such as BuildKite and GitHub Actions, and workflow orchestration in Temporal
  • Retrieve documentation and runbooks from connected knowledge sources such as Confluence, Notion, and Slack
  • Correlate findings across telemetry, code, workflows, and documentation to produce evidence-backed answers
tierzero_ask invokes the TierZero agent to process your query, which may take several minutes to complete. Clients that support MCP progress notifications will display progress updates during execution.
If you already have a TierZero chat, investigation, or artifact, use the fetch-context flow instead of starting a new investigation. Claude Code fetching existing TierZero chat context and using it to answer a follow-up question The underlying MCP tool is tierzero_fetch_context. It accepts: In Claude Code, you can invoke that flow explicitly with: 
/tierzero:fetch-context https://app.tierzero.ai/chat/c/<GlobalID>
  • Chat URLs, for example https://app.tierzero.ai/chat/c/<GlobalID>
  • Investigation URLs, for example https://app.tierzero.ai/investigations/<GlobalID>
Arguments:
  • url: A supported TierZero chat or investigation URL, or a raw GlobalID
  • include_sources: Optional boolean. Set to true to include source metadata in the response
Use tierzero_fetch_context when you want structured JSON for an existing TierZero conversation, investigation, or artifact without asking TierZero to run a new investigation.

​
Troubleshooting

​
Connection Issues

If you’re having trouble connecting:
  1. Verify your PAT: Ensure the token is valid by testing it with a direct API call
  2. Check URL: Make sure you’re using the correct URL
  3. Authorization header: Verify the header format is exactly Authorization: Bearer YOUR_TOKEN

​
Authentication Errors

If you see authentication errors:
  1. Generate a new PAT from the API keys page
  2. Update your MCP server configuration with the new token
  3. Restart Claude Code or Cursor