its just so funny to me because this has been built so many times before @jj and i even tried building it with stride.systems clawdbot is obviously different in that it's local. but the general idea is the same 1. Give full access, context, and control to your agents 2. Agents work proactively in the background for you 3. Have useful human in the loop touchpoints for approving dangerous actions there have been tons of startups that are trying to build the connective layer for ai that can not only respond in text, but actually take action even mcp touched on this idea for whatever reason, i'm not that interested in tinkering with clawdbot because it reminds me of mcp, which had a ton of hype and largely fizzled out. it feels more like productivity maxxing and config tinkering than being actually valuable the core problem of what clawdbot is solving for is inevitable. and there have been countless apps/ startups / attempts to make it work it's another experiment, but i don't think it's the final end-all solution just yet and the trouble we had with stride was that: 1. we weren't using it ourselves that much 2. we couldn't really come up with any specific, compelling use cases that resonated (with ourselves or anyone else) building a general ai tool that connects to all your stuff sounds useful, but in practice it is less so, esp. with regards to how you market it. it's hard for people to figure out how to connect all the dots, even with heavy guiding (except maybe for the n8n bros) similar fate happened to cobot.co, which was an AI todo list https://x.com/jsngr/status/1985394049402937533?s=20