I am interested in discussing what MCP adoption currently looks like in your organizations. Do you have a corporate strategy for connecting your applications to your AI through MCP? Or are you running shadow AI operations and configuring MCP yourself on the side?
At Safe, in Product and Engineering, we are predominantly using Claude. We now have MCP servers on top of most systems. For systems that do not have MCP servers, we have used FME Flow MCP to connect them. The biggest issue seems to be reliability, both Claude’s ability to pick and use the correct MCP Tool. Then the tool actually working when it tries to connect.