Tencent Cloud APM Performance Analysis

Problem

When production services show slow APIs, abnormal traces, or fluctuating metrics, engineers often have to pull APM data from multiple sources, manually correlate metrics, logs, and traces, and then identify the bottleneck. Tencent Cloud APM Performance Analysis consolidates this workflow into a single toolchain: it connects to a remote APM MCP Server through the official Tencent Cloud SDK method ApmClient.SendMCPMessage, then discovers available tools and invokes the appropriate performance analysis capability after user confirmation.

Workflow

The skill follows a confirm-before-invoke pattern:

• Check environment variables such as TENCENTCLOUD_SECRET_ID; if missing, guide configuration using the provided template without writing keys into project files.
• Use the embedded client to run ping and verify connectivity.
• Use list-tools --output json to discover available tools.
• Present tools and parameters, then confirm the target, scope, and optional arguments.
• Execute the call with call-tool --name <tool> --args '{...}'.
• Interpret the response and provide performance analysis recommendations.

During invocation, credentials are managed through shell environment variables and authenticated with Tencent Cloud’s standard API signing method TC3-HMAC-SHA256. The skill avoids hard-coding secrets into scripts, logs, or conversation; if a user provides plaintext credentials, they should be configured in the terminal and not echoed.

Boundaries

This skill fits teams that already have Tencent Cloud APM permissions and want to run diagnostics through MCP-style tooling. The environment must have network access, the correct Endpoint, and completed service role authorization; authorization failures should be resolved in the console. In mobile or restricted environments, the workflow must avoid IDE confirmation popups, must not create temporary files or run deletion commands, and should present data directly in Markdown tables, code blocks, or tree structures.