What's the difference between using a proprietary AI assistant (like Kestrel's built-in agent) versus connecting Claude directly via MCP?

Both the in-app Kestrel AI agent and the Claude connector use the exact same underlying instruction set via the same MCP server, so the underlying data returned is identical regardless of which interface you use. The key difference is scope: the Kestrel web app agent has zero other skills -- it isn't connected to the internet and can't do anything beyond what's been explicitly built for it, which also requires a user account. Claude, when connected via MCP, is just one more tool set within Claude's broader capabilities -- so it can do things the narrow in-app agent can't, like build a map of all your properties, calculate driving distances, or find nearby points of interest along a route, because those capabilities come from Claude itself, not from the MCP server. The MCP server is described as the unifying piece that guarantees consistent data whether you're working inside the Kestrel assistant or inside Claude.