“pscale MCP server kept a Neon-to-PlanetScale sync inside one agent loop”

I used PlanetScale's pscale MCP server to run a Neon-to-PlanetScale data sync, and the thing I keep coming back to is that I never had to leave the agent loop to do it. Normally a migration like this means juggling two clients: one psql/Neon session to pull sample data and read the source schema, another to poke at the destination and confirm rows actually landed. Here the MCP tooling handled the sample-data pulls and the sync checks from inside a single loop, so the agent could read from Neon and write to PlanetScale without me dropping down to a separate shell to babysit either side. My verification pass was deliberately boring, which is what I wanted. I had the agent compare index counts on the PlanetScale side against the Neon source to make sure the schema came across and nothing silently got dropped in translation. Then I inserted a test row and checked that it propagated, because counting indexes tells you the structure is there but not that writes behave. Both of those steps ran through the same MCP server I'd used to kick off the sync, so the loop that did the work also did the checking — no context-switch, no copy-pasting connection strings between terminals. What that bought me was a sync I could actually reason about end to end in one place. The agent pulled the sample data, ran the index comparison, and did the test-row insert as a continuous sequence instead of me stitching together output from three different tools. For an exploratory cross-provider move where I'm half-expecting something to be subtly off, having the read, the write, and the verify all routed through one server is genuinely the feature. Knocking off a star because this was one careful run, not a battle-tested migration of real production volume, and I was watching the sync closely the whole time rather than trusting it blind. I also can't speak to how it behaves on a large or schema-heavy database under load — my test was a controlled comparison, not a stress test. But for the specific job of letting an agent move data between Neon and PlanetScale and confirm it landed without leaving the loop, the pscale MCP server did exactly what I needed and the workflow held together. I'd reach for it again the next time I'm evaluating a move onto PlanetScale.

• No comments yet.