IMO the biggest issue with CLIs for agents is to know when the agent is allowed to type. When is the command fully proceed, and next tokens can now be generated.
This would work. We started prototyping this way and it worked well to an extend. The main issue you're going to have, is you need to have both chromium and the forwarder process to run at the same time. This makes it difficult to track what is running exactly.
If you're forwarder is running but the browser crashed, it will still appear as available on the Redis side. While if you're making the browser setting a key with expiration itself, if it crashes, it will stop appearing on redis side.
Those are the kind of stuff you get modifying the source, and that will be hard to do with dual process type of setup.
Thank you! I thought people might know it more as Google Chrome. Hence why I changed it on the HN title. But the original blog title specifically mention Chromium.