bdhcollectiveQ: Question about MCP + live browser control:
Through your MCP server, can an external AI assistant (Claude) send live, ad-hoc browser actions in the moment — e.g. "click this, type that, add this block" generated on the fly from a screenshot — and have TaskMagic execute them without a pre-built or pre-recorded flow?
Or does the MCP only let the assistant trigger flows I've already built in TaskMagic?
Put simply: can Claude be the brain deciding each step live, with TaskMagic as the hands executing it in my logged-in browser session — no recording required each time?
Deanna_TaskMagic
Jul 25, 2026A: At the moment, MCP is used to list or run flows, as well as see the run history, but we do have a feature that's currently in beta testing that allows Claude or other platforms to build browser agents via MCP. Claude wouldn't necessarily be sending browser actions for TaskMagic to execute, but Claude would then be able to send the actions it's looking to complete and turn that into a repeatable, deterministic workflow that doesn't use AI credits every single time it runs.