Investigate production issues from Cursor with the Dynatrace plugin

Dynatrace is now listed in the Cursor Marketplace as a verified plugin. One install connects Cursor to your Dynatrace environment through the Model Context Protocol (MCP) server and loads 30 Dynatrace skills, giving Cursor access to live Dynatrace data and the context to effectively analyze the data.

Figure 1: The Dynatrace plugin in the Cursor Marketplace. Browse the listing directly at cursor.com/marketplace (Cursor Marketplace, © Cursor 2026)

Connecting your environment

Connect Cursor to your Dynatrace environment using a platform token. You can create one in your Dynatrace environment under platform tokens. Once configured, Cursor can access the live Dynatrace data through the plugin, which manages the connection between Cursor and your environment. We're already working on a simpler authentication flow in a forthcoming update.

What does a Cursor Marketplace listing include

A Cursor plugin can bundle more than a single MCP connection. For example, one plugin can ship MCP servers, agent skills, rules, hooks, and commands together, including dt-dql-essentials for querying syntax and patterns, dt-app-dashboards and dt-app-notebooks for building platform assets, and dt-alerting for anomaly detector setup.

An agent with MCP access can reach your data, but it might still write queries against fields that don't exist unless it knows how Dynatrace is organized. The skills included in the Dynatrace plugin for Cursor carry that knowledge and load only when a task calls for them, so installing 30 of them doesn't crowd the context window. These are the same skills we released as Dynatrace for AI, now packaged for a one-click install.

Teams and Enterprise admins can distribute the plugin through a private team marketplace, so a platform team decides how developers connect to Dynatrace.

Figure 2: The plugin listing shows one MCP server and 30 Dynatrace skills (Cursor Marketplace, © Cursor 2026).

How does a developer debug a failing service in Cursor?

Say payment-api starts returning 5xx errors after your last deploy, and you already have the code open.

1. Ask why payment-api is failing, in plain language.
2. The agent pulls the root cause into the session along with the related logs, metrics, traces, and CPU and memory profiling data from your live environment.
3. You narrow by impact, filter to the service, and ask for remediation suggestions.
4. You fix the code without leaving the editor.

You did not have to design the query yourself and you never had to guess where to look. The root cause information is grounded in Dynatrace causal analysis and Smartscape™ topology data, helping the agent explain why an issue occurred instead of relying on inferred correlations.

The screenshot below shows this in practice: the agent pulling live root cause data directly into a Cursor session.

Figure 3: Cursor agent session showing live Dynatrace production data pulled into the editor (Cursor Marketplace, © Cursor 2026).

The answer comes from your actual environment, not a documentation summary or a pre-cached result.

Which questions should a developer ask?

Cursor becomes another surface for the same conversational analysis you run in Dynatrace Assist. Areas with the potential for high payoff at the keyboard:

• Incident analysis. Root cause of a problem ID, log patterns on a Kubernetes workload, service-to-service dependencies, and the real user impact of a regression you shipped.
• Security. "Am I vulnerable to a specific CVE affecting workloads running in production?" answered against the workloads actually running in production, not against a dependency manifest.
• Optimization and tuning. Oversized hosts, database calls that are slow at p99, and forecast headroom before you scale.

Developers can benefit from answers that are grounded in your environment rather than in a documentation summary.

Run platform tasks from the terminal

The plugin covers query and analysis out of the box. For write operations, such as creating dashboards, triggering workflows, and executing DQL mutations, pair it with dtctl, Dynatrace's open source CLI. Download it from dynatrace-oss/dtctl and run it from the Cursor CLI or your terminal environment alongside the MCP server.

Consumption is priced on data, not per user, so onboarding a second or a fiftieth developer into this workflow doesn't change what you're billed.

Dynatrace, Grail, Smartscape, and related marks are trademarks of the Dynatrace group of companies. All other trademarks are property of their respective owners.