Requested on Discord.
Add an MCP tool that moves data from one connection to another, e.g. copying rows from prod down to dev or local for testing. Today this means export results, then import results, by hand each time.
Rough shape:
move_data(source_connection, dest_connection, query, dest_table)
- Runs the query on the source and inserts the rows into the destination table.
- Does the work behind the scenes and returns a status/verification summary at the end. It must not return the rows themselves, which would blow up the LLM context window.
- Needs a flag for handling duplicate IDs (skip/replace).
- Expected volume: up to a few hundred thousand rows, one table at a time... but no reason someone does not try it on millions.
Open question: how to report progress to the LLM while the tool runs on larger copies. MCP progress notifications may cover this, needs investigation.
Requested on Discord.
Add an MCP tool that moves data from one connection to another, e.g. copying rows from prod down to dev or local for testing. Today this means export results, then import results, by hand each time.
Rough shape:
move_data(source_connection, dest_connection, query, dest_table)Open question: how to report progress to the LLM while the tool runs on larger copies. MCP progress notifications may cover this, needs investigation.