--load: scenario file with a weighted mix of tools, resources/read and prompts/get #2839

Which client?

All / shared core

The problem

Real traffic is not one tool called with fixed arguments. --load can only drive a single tools/call, so a server's behaviour under a realistic mix (several tools, resources/read, prompts/get) cannot be measured.

Solution you have in mind (optional)

--load-scenario scenario.json listing steps with a weight each, e.g. [{"method":"tools/call","tool":"search","args":{...},"weight":7},{"method":"resources/read","uri":"...","weight":3}]. The summary breaks down counts, errors and latency per step plus the total.

Alternatives or workarounds you have tried

Running several --load processes in parallel, one per tool — which multiplies connections and gives no combined view.

Already built it locally?

No response

Before you submit

• I searched existing issues and this is not a duplicate.
• This is a request for the Inspector itself, not for the MCP specification or an SDK.