Problem statement
Currently, when a tool is invoked during an MCP server run (e.g., by an agent, user query, or workflow), there is zero built-in visibility into what happened. Developers have no straightforward way to observe:
- Which specific tool was called
- When the tool execution started and finished
- How long the tool took to complete
- Whether the tool call succeeded or failed
- Any basic error message if it failed
Proposed solution
Emit structured logs for every tool/prompt/resource execution:
[tool] Executing 'get-weather' | input: { "location": "Buenos Aires" }
[tool] 'get-weather' completed | duration: 320ms | success
This should be enable/disable from the xmcp.config.ts
Package
xmcp (core framework)
Problem statement
Currently, when a tool is invoked during an MCP server run (e.g., by an agent, user query, or workflow), there is zero built-in visibility into what happened. Developers have no straightforward way to observe:
Proposed solution
Emit structured logs for every tool/prompt/resource execution:
This should be enable/disable from the
xmcp.config.tsPackage
xmcp (core framework)