MCP Doesn't Suck. Your Agent Does. — Jan Čurn, Apify

The Model Context Protocol (MCP) faces significant criticism for causing context bloat and inefficiency, yet these issues stem from poor implementation rather than the protocol’s design. Effective agent-tool interaction requires strategies like progressive tool discovery and "CodeMode," where models treat tools as executable code rather than static context. While command-line interfaces (CLIs) naturally excel at these tasks, they lack a standardized remote access protocol. MCPC, a universal CLI client, resolves this by wrapping MCP functionality into a lightweight, standard-compliant interface. By supporting asynchronous tasks, JSON-based composition, and secure credential management, MCPC allows developers to leverage the full power of MCP without sacrificing context window efficiency. This approach demonstrates that combining the standardized connectivity of MCP with the operational efficiency of CLI tools provides the optimal architecture for modern AI agents.