MCP Explained
An interactive explainer site for the Model Context Protocol (MCP) — a standard for letting AI agents call external tools and services — is drawing mixed reactions. Many find the multipane, animated format confusing, calling it “AI slop” and arguing that clear, text‑based primers or simple code examples would teach the concept better, especially for beginners. Beyond the explainer’s design, commenters question whether MCP adds enough value over existing APIs and CLIs, with skeptics seeing it as unnecessary bloat and proponents highlighting its role in secure, auditable tool access for enterprises and non‑technical users.
Reception of the MCP explainer site
- Many find the interactive explainer “too dense,” noisy, and visually overwhelming (multiple panes updating at once, text changing too fast).
- Several call it “AI slop” or UX-first “vibe dashboards” that look slick but don’t actually clarify concepts.
- A minority defend it as one of the better MCP explainers so far and note the author has already simplified it in response to feedback.
Documentation, explainers, and LLMs
- Strong sentiment that modern docs are often bad and LLMs are worsening things by encouraging “just ask AI” instead of writing real guides.
- Multiple comments distinguish between:
- Guides/introductions vs. reference docs.
- Explainers vs. proper documentation.
- Some argue that the best docs are written by people who just learned the tech, then reviewed by experts.
- Others say LLM-based personal explainers are fine for self-learning but not worth sharing.
What MCP is and protocol nitpicks
- Confusion and debate over whether MCP is “basically REST,” JSON-RPC, or something distinct.
- Several emphasize that MCP uses JSON-RPC and is explicitly “not REST,” while others focus on practical similarities (stateless requests, JSON payloads).
- Some liken MCP to older middleware standards (CORBA, COM+, SOAP, LSP).
Usefulness vs. redundancy of MCP
- Skeptics: MCP is “nothing burger,” “garbage standard,” or XML-style bloat; CLIs + HTTP APIs + code execution are seen as enough, especially as models get better at using arbitrary APIs.
- Supporters: MCP standardizes how agents discover and call tools, especially when agents don’t have shell access or local interpreters.
- Several report strong success using MCP to expose domain-specific systems (scientific data, admin portals, home automation, SaaS tools) to non‑technical or lightly technical users.
Security, auth, and enterprise angle
- Repeated claim that MCP’s real value is in controlled, auditable access: OAuth integration, read-only scopes, and avoiding giving agents raw shell or database credentials.
- Seen as especially relevant in regulated or enterprise environments and for third‑party tools.
Composability and ergonomics
- Critics say MCP tools aren’t naturally composable like shell/CLI pipelines; streaming, signaling, and composition often depend on the harness and are easy to get wrong.
- Some report better results when MCP servers expose just a couple of “execute script”/“get help” tools rather than many granular endpoints.
- There’s a sense MCP is most useful when deeply integrated and mostly invisible; otherwise, a hassle.