thinking about agentic ai… what if before it started a task, it could look for mcp tools it might need to complete a task, then it would install those tools by itself before it started. With a large enough tools registry, an agent could accomplish basically anything by itself
Where are you finding out how MCPs work?
Please Login to reply.
Had to look at the raw schema and manually query jsonrpc on mcp endpoints to get a feel for how it works
Ugh. I found one which lets me access the shell. That works. This is all oddly both under documented and really powerful
