About the Dash0 MCP ServerRelease historyThis feature was introduced by Dash0 MCP Server on June 11, 2025.

The Dash0 MCP Server lets AI assistants such as Claude, Claude Code, and Cursor query your Dash0 observability data. It uses the Model Context Protocol (MCP), an open standard for connecting AI assistants to tools and data. Ask a question in natural language, and the assistant calls Dash0 tools to read services, metrics, logs, traces, synthetic checks, SLOs, and dashboards.

What the Server Provides

• Read-only telemetry tools: Look up the service catalog, run PromQL and SQL queries, search logs and spans, open traces, and list failed checks, synthetic checks, SLOs, dashboards, and websites. These tools only read data. They never create, change, or delete anything in Dash0.
• Agent0 delegation: The runTask tool hands an open-ended question to Agent0, which investigates across metrics, logs, and traces and reports back. Each task creates an Agent0 thread in your organization, which you can open and continue in Dash0. That thread is the only thing the server creates, and Agent0 itself only reads over MCP. Delegation consumes Agent0 credits, and an organization Admin can turn it off. See Delegate to Agent0 via MCP.
• Links back to Dash0: Most results include a link that opens the same view in the Dash0 app.

Before You Connect

• A Dash0 account: You need a Dash0 user who is a member of an organization. The server needs no extra setup in Dash0. Every member can connect.
• Your region's MCP endpoint: Each Dash0 region has its own endpoint. Copy yours from Settings → Endpoints under MCP, or pick it from the table below.
• A Claude plan that supports connectors: On Claude Team and Enterprise plans, an owner may need to allow the connector for your organization before members can add it.

Dash0 hosts the server and it uses the streamable HTTP transport, so there is nothing to install or run on your machine.

Connect from Claude's Directory

1. In Claude, open Settings → Connectors and click Browse connectors.
2. Search for Dash0 and click Connect.
3. If Claude asks for a region or URL, choose the one that matches your Dash0 region. Check it under Settings → Endpoints in Dash0.
4. A Dash0 window opens. Sign in, choose the organization you want to connect, and click Allow access.

Claude now lists the Dash0 tools under the connector. In a conversation, open the Search and tools menu to turn individual tools on or off.

Connect as a Custom Connector

If the Dash0 listing isn't available in your region yet, add the server as a custom connector:

1. In Claude, open Settings → Connectors and click Add custom connector.
2. Enter a name such as Dash0 and paste your region's MCP endpoint URL.
3. Click Add, then Connect, and complete the Dash0 sign-in as above.

Connect from Claude Code and Other ClientsRelease historyThis feature was introduced by OAuth Authentication for MCP on April 16, 2026.

Claude Code, Codex, Cursor, Windsurf, and other MCP clients connect to the same endpoint. Each client opens the Dash0 sign-in in your browser on first use and refreshes the token after that.

Claude Code:

claude mcp add --transport http --scope user dash0 https://api.<region>.<cloud>.dash0.com/mcp

--scope user makes the server available in all your projects. Use local or project to limit it. Run /mcp in Claude Code and select dash0 to sign in.

Codex:

codex mcp add dash0 --url https://api.<region>.<cloud>.dash0.com/mcp
codex mcp login dash0

Cursor, Windsurf, and other clients: Add a remote MCP server with your region's endpoint URL in the client's MCP settings, and leave out any Authorization header so the client uses the Dash0 sign-in.

For more setup snippets, see the Dash0 MCP Server page in the Integration Hub.

Let Your Coding Agent Set It Up

Paste this prompt into Claude Code, Codex, Cursor, or another coding agent. It asks for your region, registers the server, and checks the connection:

Connect this assistant to the Dash0 MCP server.

1. Ask me which Dash0 region my organization uses, unless I already said so.
   I can look it up in Dash0 under Settings → Endpoints → MCP. The endpoints are:
   - AWS Europe (Ireland):              https://api.eu-west-1.aws.dash0.com/mcp
   - AWS Europe (Frankfurt):            https://api.eu-central-1.aws.dash0.com/mcp
   - Google Cloud Europe (Netherlands): https://api.europe-west4.gcp.dash0.com/mcp
   - AWS US East (N. Virginia):         https://api.us-east-1.aws.dash0.com/mcp
   - AWS US West (Oregon):              https://api.us-west-2.aws.dash0.com/mcp
2. Register a remote MCP server named "dash0" with that URL over streamable HTTP.
   Use OAuth: don't add an Authorization header.
   - Claude Code: claude mcp add --transport http --scope user dash0 <URL>
   - Codex: codex mcp add dash0 --url <URL>, then codex mcp login dash0
   - Any other client: add a remote MCP server entry with this URL to its MCP configuration.
3. Tell me how to finish signing in. In Claude Code, I run /mcp and select dash0.
   I sign in to Dash0 in the browser and choose my organization.
4. Once connected, call listDatasets and show me my datasets. Then list the
   services in the dataset I pick.

Connect with an Auth Token

For CI/CD pipelines, scripts, and other places without a browser, authenticate with a Dash0 auth token instead of OAuth. The token needs read access to the datasets you want to query. An ingest-only token can't read data.

In Claude Code:

claude mcp add --transport http dash0 https://api.<region>.<cloud>.dash0.com/mcp \
  --header "Authorization: Bearer <DASH0_AUTH_TOKEN>"

In clients configured with an mcpServers JSON file that accepts remote servers, such as Cursor:

{
  "mcpServers": {
    "dash0": {
      "url": "https://api.<region>.<cloud>.dash0.com/mcp",
      "headers": {
        "Authorization": "Bearer <DASH0_AUTH_TOKEN>"
      }
    }
  }
}

For clients that only run local MCP servers, use mcp-remote as a bridge:

{
  "mcpServers": {
    "dash0": {
      "command": "npx",
      "args": ["mcp-remote", "https://api.<region>.<cloud>.dash0.com/mcp", "--header", "Authorization:${AUTH_HEADER}"],
      "env": {
        "AUTH_HEADER": "Bearer <DASH0_AUTH_TOKEN>"
      }
    }
  }
}

A token connection has no person behind it, so it uses the token's permissions instead of a user's, and Agent0 runs it at the Dash0 default model level.

How Datasets and Permissions Apply

• You act as yourself: The connection uses an OAuth token that acts on your behalf in the organization you chose at sign-in. It has the same permissions as your Dash0 user, bounded by your organization role and your dataset access. It can never reach data you can't see in the Dash0 app.
• Permissions are checked on every call: If an Admin removes your access to a dataset, or removes you from the organization, the connection loses that access immediately.
• One organization per connection: To query a second organization, connect again and choose that organization at sign-in.
• Every query names a dataset: Most tools require a dataset argument, so each call reads one dataset. The assistant calls listDatasets to discover the datasets you can access. Name the dataset in your prompt, for example "in the production dataset", to choose which one it reads.
• Agent0 runs read-only: Investigations started through runTask use a read-only credential. Over MCP, Agent0 can't create dashboards or check rules, open pull requests, or write to any connected system.
• Short-lived tokens: Access tokens expire after 15 minutes and refresh automatically. Every authorization and token exchange appears in your organization's audit log.

Disconnect

• From Claude: Open Settings → Connectors, select Dash0, and click Disconnect.
• From Dash0: Open your user profile, go to Applications, and click Revoke next to the client. The client loses access to that organization immediately.
• For the whole organization's Agent0 use: An Admin can turn off Agent0 via MCP under Settings → Agent0. This removes runTask and the other Agent0 tools for every member. The read-only telemetry tools stay available.

Example Prompts

Replace the dataset and service names with ones from your own organization.

Troubleshooting

The sign-in window says the request is no longer valid

The authorization request expired or was already used. Close the window and click Connect in Claude again. Requests expire after 10 minutes.

Claude connects, but tools return no data

• Wrong region: Each region is a separate Dash0 deployment. Check that the connector URL matches the region under Settings → Endpoints.
• Wrong organization: The connection reads only the organization you chose at sign-in. Disconnect and connect again to choose another.
• Wrong dataset: Ask the assistant which datasets it can see, then name the one that holds your telemetry in your prompt.

Tools fail with an authorization error

Your access token couldn't be refreshed, or the client was revoked. Disconnect the connector in Claude and connect again. If it still fails, check with an Admin that you're still a member of the organization and have access to the dataset.

runTask or the other Agent0 tools are missing

An Admin turned off Agent0 via MCP under Settings → Agent0, or Agent0 isn't enabled for your organization. The read-only telemetry tools keep working. Reconnect after an Admin turns it back on so Claude reloads the tool list.

runTask returns a thread link instead of an answer

The investigation took longer than the wait time and keeps running in Dash0. Ask the assistant to wait for the result, which calls waitForTask, or open the thread link in Dash0.

Requests are rejected as rate limited

The server limits how many tool calls an organization can make in a short time, across all of its members' connections. Wait a moment and retry, or ask the assistant to make fewer, broader queries.

For anything else, contact Dash0 support. Report security issues as described in Dash0's security.txt.

Further Reading

• Delegate to Agent0 via MCP — Hand whole investigations to Agent0, its cost, and how to turn it off
• Integration Hub — Setup snippets for each AI assistant
• Privacy Policy — How Dash0 collects, uses, and retains data
• Use AI in Dash0 — Overview of Dash0's AI capabilities