Dynamic Data Source Selection

Prerequisites

• Search the current open issues

What are you trying to do that currently feels hard or impossible?

Currently, if i have to setup tools, they are binded to one data source. For ex, for blocking for 10 MySQL servers, i would have 10 different tools with the exact same query. Can the servername be passed as a header optionally during the call to call the corresponding mcp tool on a dynamic server?

Suggested Solution(s)

As an example, i have below for sql server which is a global tool to check for blocking and the target which is already configured in the data sources, can just be passed as a header

tools:
detect-blocking:
engine: mssql
timeout: 10
statement: |
SELECT *
FROM sys.dm_exec_requests
WHERE blocking_session_id <> 0;
description: |
Get blocking sessions from SQL Server

{
"target_source": "xxxx"
}

Alternatives Considered

No response

Additional Details

No response

No files or tests are named. Start by tracing how tools are bound to configured data sources and how a call could provide the requested target source, then define the routing and validation behavior; done means one tool can select among configured database sources without duplicating the tool definition.