Vibe coding tip: Run playwright MCP and other token-hungry tools in a sub agent to preserve your main context window
Vibe coding / augmented engineering tip of the week pt. 6: Run playwright MCP and other high token usage MCPs or tasks in a sub agent in claude code so it doesn't eat your main context window nearly as fast.
Huge thanks to Anthony Panozzo for tipping me off to this one - I think the playwright MCP is really useful for testing changes and verifying things, but it burns through tokens at a crazy clip.
I'm also finding it useful when you're running a test suite that spits out a lot of logs or task output, etc. to have it run that in a subagent and pull out the key findings, then take the correct next action for your current goal.
Way fewer compactions.
Modern models are good enough that one-sentence prompts often get 85% of the value with 1% of the effort. The barrier to getting started is lower than people think.
How connecting the right Model Context Protocol servers in AI coding tools dramatically improves output quality
Persona-style agents are fun to name, but juggling Claude Code, Codex, ChatGPT and more calls for a single interface. Right now the answer might be to build it myself: a filterable Kanban board with a direct line to each agent.
Get More Like This
Follow along as I build and share what I learn
Found this helpful? Share it with your network!