jawuilp

I usually use MCP when I want to make sure the agent on duty doesn't get confused while performing an action. For example, I use Supabase's MCP to search for data in the database, and I don't want the agent to use the CLI.

Interesting, I've had almost the opposite experience. I ended up going back to a code editor because CLI agents tend to scatter my attention quite a bit, especially when working on apps with lots of different flows. I prefer a more controlled and focused environment.

Do you also delegate small changes to agents, or do you still make those manually?

I get what you mean. I don't think autocomplete has changed radically either, which isn't necessarily a bad thing.

What autocomplete/editor were you using when you felt the productivity gain was mostly a placebo? I've mostly used it in VS Code and a bit in Antigravity.