Dash0 MCP Server

Overview

The Dash0 MCP Server is the official Model Context Protocol (MCP) server for Dash0. It enables AI assistants to interact with your observability data, allowing them to troubleshoot issues and read services, metrics, logs, traces, failed checks, synthetic checks, SLOs, dashboards, and websites. All of its tools are read-only.

The MCP server is a "remote" MCP server, cloud-hosted by Dash0. It uses the Streamable HTTP transport, so there is no local process to install or manage. AI assistants connect to it remotely using OAuth 2.0 or a Dash0 auth token.

The MCP server can also delegate read-only investigations to Agent0, Dash0's AI observability assistant. Agent0 investigates and returns structured findings with a recommended fix, but it does not apply the change itself. A full investigation consumes roughly 1-2 Agent0 credits.

Use Cases

• Investigate production incidents by querying logs, traces, and metrics from within your AI assistant
• Analyze service health and performance without leaving your development environment
• Explore OpenTelemetry resources and their relationships through natural language
• Automate observability workflows by letting AI agents retrieve and correlate telemetry data
• Delegate investigations to Agent0 for deeper root-cause analysis and a recommended fix

Setup

The Dash0 MCP Server uses the Streamable HTTP transport, running as a cloud-hosted remote service. AI assistants can authenticate using either OAuth 2.0 (recommended for interactive AI tools like Claude Code, Codex, Cursor, and Windsurf) or a Dash0 auth token (recommended for CI/CD and automated workflows).

Prerequisites

Before configuring the MCP server, ensure:

• You are a user of a Dash0 organization. Every user can connect, and no extra setup is required
• You know your MCP endpoint URL — find it under Settings → Endpoints in Dash0, or pick the one for your region from the table below
• For bearer token auth: an auth token with the All permissions or Reading permission. Tokens with the Ingesting or Signal Control edge permission do not work with the MCP server

MCP Endpoints

The MCP endpoint URL has the form https://api.<region>.<cloud>.dash0.com/mcp. Use the <region> and <cloud> values that match the region of your Dash0 organization:

For example, an organization in AWS Europe (Ireland) uses the eu-west-1 region on aws.

OAuth 2.0 (Recommended)

OAuth 2.0 lets AI assistants authenticate without static API tokens. Dynamic client registration, browser-based consent, and token refresh are handled automatically on first use.

Claude Code

claude mcp add dash0 {{endpoint_mcp}} --transport http

Add --scope to control where the server is registered, choosing one of local, user, or project (for example, --scope user makes it available across all your projects). See the Claude Code MCP scopes documentation for details.

On first use, Claude Code opens a browser window where you log in to Dash0 and grant consent. Access tokens expire after 15 minutes and refresh automatically.

Claude (claude.ai and Claude Desktop)

1. Open Settings → Connectors and choose Add custom connector.
2. Enter Dash0 as the name and paste the MCP endpoint URL for your region.
3. Choose Connect, sign in to Dash0, and select the organization to connect.

Codex

1. Add the Dash0 MCP server:

   bash

   codex mcp add dash0 --url {{endpoint_mcp}}

2. Authenticate with Dash0:

   bash

   codex mcp login dash0

Codex opens a browser window where you log in to Dash0 and grant consent. You can review and revoke connected applications under User Settings → Applications in Dash0.

Cursor, Windsurf, and other MCP-compatible clients

Cursor, Windsurf, and other clients that support remote MCP servers with OAuth can connect using the same MCP endpoint URL. Add the server through your client's MCP configuration UI or mcp add command, omit the Authorization header, and the client will walk you through the OAuth consent flow on first connection.

Bearer Token (for CI/CD and Automation)

For headless environments, scripts, and CI/CD pipelines, configure the server with a static Dash0 auth token.

Claude Code

claude mcp add --transport http dash0 {{endpoint_mcp}} --header "Authorization: Bearer {{token}}"

Cursor and other clients with mcpServers JSON

Clients whose mcpServers JSON configuration accepts remote servers, such as Cursor, can connect directly using the Streamable HTTP transport:

{
  "mcpServers": {
    "dash0": {
      "url": "{{endpoint_mcp}}",
      "headers": {
        "Authorization": "Bearer {{token}}"
      }
    }
  }
}

Other AI Assistants

For AI assistants that do not yet support remote MCP servers natively, you can use mcp-remote as a bridge:

{
  "mcpServers": {
    "dash0": {
      "command": "npx",
      "args": ["mcp-remote", "{{endpoint_mcp}}", "--header", "Authorization:${AUTH_HEADER}"],
      "env": {
        "AUTH_HEADER": "Bearer {{token}}"
      }
    }
  }
}

Billing

Read-only telemetry queries over MCP are free. Invoking Agent0 to run an investigation consumes Dash0 credits, because Agent0 performs a multi-step investigation on your behalf — the same as a run started from the Dash0 UI.

An administrator controls whether Agent0 is available over MCP, under Settings → AI. Turn Agent0 via MCP off there to stop anyone in your organization from spending credits this way.

Verification

1. Open your AI assistant and ask it to list available Dash0 tools
2. Try a simple query such as "Show me the services in Dash0"
3. Verify that the assistant can retrieve metrics, logs, or traces from your organization

References

• Dash0 MCP Server documentation
• Model Context Protocol Specification