2 comments

[ 1.0 ms ] story [ 8.9 ms ] thread
Hi HN, I built this because my coding agent subscription quota was running out too quickly.

I already had access to web-based AI with much more usage, so I started wondering: why not let the web AI do the thinking, and let the coding agent do the actual work?

MCP made this possible.

AgentBridge connects the two: the Brain can inspect the local project and create a plan, while the local coding agent executes the changes and runs tests.

It started as a way to make better use of my AI subscriptions, but I think the Brain/Executor separation could be useful beyond that.

Would love to hear what you think.