Dynatrace MCP server: remote setup, tools and limits

Dynatrace hosts its MCP server inside your SaaS environment. Its 20 documented tools read or analyze data, and a DQL query returns at most 1,000 records.

Checked 7 October 2026 · from Dynatrace's MCP documentation and Hub listing, the retired repository, and that server's own tools/list response

One of 91 guides to monitoring and developer tools. Each covers what an AI coding agent can get out of the tool through MCP, the protocol that lets the agent query the tool directly.

What is the Dynatrace MCP server?

It is Dynatrace's server for the Model Context Protocol. The agent gains tools for querying Grail, reading problems and vulnerabilities, and running Davis AI analyzers. Davis AI is Dynatrace's own causal and predictive AI.

Grail is Dynatrace's data store and DQL is the query language over it. Grail Query Agent writes a DQL query from English, and DQL Explanation Agent explains one. Data Analysis Agent runs the query.

Dynatrace offers three routes. The Remote MCP Server is the MCP route. `dtctl` is a command-line tool that Dynatrace pairs with it for management tasks. The open-source server is archived and receives no further updates.

Dynatrace has two MCP servers, and one is current:

dtctl is the third first-party route. It is a command-line tool, so it appears under the alternatives below.

Dynatrace covers production observability broadly and reaches part of the way into release through deployment events. It reaches nothing in code review, agent observability or cost.

In practice, Dynatrace answers questions about the running system in depth, and nothing about the code that produced it or the agents you operate. An agent connected here can find a problem, query the logs around it and run a Davis analyzer on the metric. It cannot see the pull request that caused it, so this connector pairs with a source-control one.

What can an AI agent do with Dynatrace?

Tool names follow Dynatrace's tool reference, read 7 October 2026. The last row is the main change from the retired open-source server, which could also send email and Slack messages.

What the tool definitions cost in context

A client that loads every tool definition at the start of a session spends this much context before the first question. Claude Code defers definitions by default through tool search, and loads a tool's definition when the agent needs it. The method below gives the version and date measured.

A context window is the amount of text a model can hold at once. A client that loads every tool definition upfront sends each tool's name, description and input schema to the model when the session starts. That list is charged whether or not a single tool is called.

The figures below were measured on the retired open-source server on 19 August 2026. Its 18 tools cost about 3,909 tokens. The Remote MCP Server answers tools/list only inside an authenticated environment, so its 20 tools are unmeasured.

Encodings: GPT-5.6, GPT-5.5 o200k_base; GPT-4, GPT-3.5 cl100k_base; Claude Haiku 4.5 to Opus 5 count_tokens. Counted 19 August 2026 from tools/list on the retired @dynatrace-oss/dynatrace-mcp-server package, at its final version 2.1.2. It was run locally with a placeholder environment and token.

Registration is static, so no Dynatrace tenant was involved. Figures cover name, description and input schema. The two models in column one share the o200k_base encoding, so they get one count. The spread is formatting, about 31%. Reasoning tokens are output, spent per request. This count is billed as input whenever the client loads it.

The Claude figures are measured rather than encoded. Anthropic publishes no local tokenizer. The definitions were sent to the API's count_tokens endpoint, and a no-tools baseline was subtracted. The range covers two tokenizer generations rather than two serializations. Haiku 4.5 and Opus 4.6 sit at the low end. Opus 5 and Sonnet 5 sit at the high end. The gap between them is about a third.

Dynatrace MCP server limits: where answers come back incomplete

The first two limits below return answers that look complete. The third applies to setups still running the retired open-source server.

How to configure Dynatrace MCP for agents

1. 1

   Use the Remote MCP Server

   Dynatrace hosts it in your SaaS environment and supports it officially. The open-source server is archived, with v2.1.2 as its final release.

2. 2

   Choose how you authenticate

   Dynatrace recommends a platform token. A client stops working when that token expires. A confidential OAuth client with its secret lets the client refresh tokens automatically.

3. 3

   Grant a scope for each data type

   A tool runs within both the token's scopes and the user's permissions. Add storage:logs:read, storage:metrics:read and the others for each data type the agent should query.

4. 4

   Plan for the 1,000-record cap

   Data Analysis Agent returns at most 1,000 records per query. Ask for aggregates in DQL when a question spans more records than that.

5. 5

   Add dtctl for management tasks

   Dynatrace pairs its command-line tool with the remote server. It handles workflows, dashboards and other changes the MCP tools leave out.

Paste this into your agent’s instructions

Two Dynatrace limits look like a complete answer. One is a query cut off at 1,000 records. The other is an empty result from a token that lacks a scope. The block makes the agent separate them.

When you answer using Dynatrace MCP tools, state:
- The DQL you ran and the time range it covered.
- Whether the result held 1,000 records. That is the tool's
  cap, so the full result may be larger.
- Which data types you queried. A token that lacks a
  storage scope returns nothing for that type.
Report an empty result as "none returned under the
constraints above".
Write your answer in ASD-STE100 Simplified Technical
English. Use short sentences with one idea in each.

Do you need the MCP server at all?

Partly. The Dynatrace platform APIs cover Grail queries and problem lists, and a coding agent can call them directly with a platform token.

dtctl is the other first-party route. It is a command-line tool for Dynatrace SaaS, built for people and agents, and ships its own agent skill. It also manages workflows and dashboards, which the MCP tools can only find.

The MCP server adds the DQL authoring tools and the Davis analyzers as ready-made tools, with Dynatrace's own descriptions. An agent can use them before it knows DQL.

Use the Remote MCP Server for questions about the running system. Add dtctl when the agent also has to change workflows or dashboards.

Dynatrace MCP server setup

The Remote MCP Server runs inside your Dynatrace SaaS environment, so there is nothing to install. The endpoint is https://<environment-id>.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp.

Every request needs a bearer token from a platform token or a confidential OAuth client. Both the user and the token need mcp-gateway:servers:invoke and mcp-gateway:servers:read, plus the scopes each tool lists.

export DT_ENVIRONMENT=https://abc12345.apps.dynatrace.com
export DT_PLATFORM_TOKEN=<your-platform-token>
claude plugin install dynatrace@claude-plugins-official

{"mcpServers":{"dynatrace":{
  "url":"https://abc12345.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp",
  "headers":{"Authorization":"Bearer <your-platform-token>"}}}}

[mcp_servers.dynatrace]
url = "https://abc12345.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp"
bearer_token_env_var = "DT_PLATFORM_TOKEN"

Every other client

Each block below is the configuration for one client, with the file path and the key that client expects.

{"mcpServers":{"dynatrace":{"command":"npx",
  "args":["-y","mcp-remote",
    "https://abc12345.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp",
    "--header","Authorization:${AUTH_HEADER}"],
  "env":{"AUTH_HEADER":"Bearer <your-platform-token>"}}}}

{"mcpServers":{"dynatrace":{
  "httpUrl":"https://abc12345.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp",
  "headers":{"Authorization":"Bearer <your-platform-token>"}}}}

{"servers":{"dynatrace-mcp":{
  "url":"https://abc12345.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp",
  "headers":{"Authorization":"Bearer <your-platform-token>"}}}}

{"mcpServers":{"dynatrace":{"type":"http",
  "url":"https://abc12345.apps.dynatrace.com/platform-reserved/mcp-gateway/v0.1/servers/dynatrace-mcp/mcp",
  "headers":{"Authorization":"Bearer <your-platform-token>"}}}}

claude mcp add-json "dynatrace" '{"command":"npx",
  "args":["-y","@dynatrace-oss/dynatrace-mcp-server"],
  "env":{"DT_ENVIRONMENT":"https://abc12345.apps.dynatrace.com"}}'

DT_ENVIRONMENT=https://abc12345.apps.dynatrace.com \
DT_PLATFORM_TOKEN=<your-token> \
DT_GRAIL_QUERY_BUDGET_GB=10 \
npx -y @dynatrace-oss/dynatrace-mcp-server

Read-only access, permissions and security

• Every documented tool reads or analyzes

  The Remote MCP Server documents query, problem, entity, security and analyzer tools. Its reference lists nothing that sends email, posts to Slack or ingests events.

• The token carries your permissions

  A tool runs within both the token's scopes and the user's permissions. Scope the token to the data types the agent needs.

• Keep the token out of shared files

  A platform token is a bearer credential. Load it from a variable, as the Dynatrace plugin does with DT_PLATFORM_TOKEN.

• The retired server could act

  Four of its 18 tools sent email, posted to Slack, ingested events or reset the query budget. A setup still running it needs those four off the allowlist.

Troubleshooting

The client cannot connect to the endpoint

Every request needs a bearer token. Both the user and the token need mcp-gateway:servers:invoke and mcp-gateway:servers:read.

A query returns no logs

The token or the user likely lacks storage:logs:read. Dynatrace notes that both need it before any logs appear. Other data types need their own storage scopes.

The connection worked yesterday and fails today

The bearer token has probably expired. VS Code and similar clients keep sending an expired token until you replace it. A confidential OAuth client with its secret refreshes automatically.

A result holds exactly 1,000 records

That is the cap. Data Analysis Agent limits Grail responses to 1,000 records. Aggregate in DQL or narrow the time range.

An unexpected email or Slack message went out

That came from the retired open-source server, where send_email and send_slack_message are on by default. Remove both from the allowlist, or move to the Remote MCP Server.

Dynatrace MCP server: Reference

What engineers report

Dynatrace hosts the Remote MCP Server, so its record is Dynatrace's own documentation and changelog. The public repository belongs to the retired server, with 137 stars as of 7 October 2026.

“MCP Server starts, but copilot never uses the MCP server. Any prompts about the exposed tools bring nothing from the MCP server.”

Repository facts from the GitHub API, retrieved 7 October 2026. Remote tool names and states from Dynatrace's tool reference, read 7 October 2026. The retired server's tool counts come from its own tools/list response on 19 August 2026.

Should you connect an agent to Dynatrace?

Yes. Best for dynatrace SaaS teams connecting the Remote MCP Server with a scoped token, plus dtctl for workflow and dashboard changes.

Strongest fit

Dynatrace SaaS teams who want an agent querying Grail, reading problems and security findings, and running Davis analyzers.

Main advantage

Hosted and supported by Dynatrace, with DQL authoring tools, and every documented tool limited to reading or analysis.

Main weakness

Query results stop at 1,000 records, and the DQL authoring and help tools are still Early Access.

Operational risk

A token that lacks a data scope returns empty results that read like real answers. Bearer tokens in a client config also expire.

Frequently asked questions