How to tell if your MCP server is actually good
Blog post from Prefect
FastMCP 4 adopts the stateless Streamable HTTP MCP protocol, allowing servers to scale horizontally without session-based sticky routing, while requiring clients to provide all necessary context with each request. The discussion argues that MCP servers should not mirror REST APIs or replace skills and CLIs, but instead provide curated, outcome-focused interfaces that combine underlying operations into concise tools for agents. Effective MCP design emphasizes minimizing tool and token footprint, using middleware to enforce data filtering, authentication, and governance, and considering code mode for exceptionally large API surfaces. Skills can complement MCP servers through plugins that package guidance, dependencies, and alternative workflows such as CLI-based mutations. The speakers also stress writing outcome-based evaluations from real support cases, running them in CI to validate server changes across models, and deploying servers with durable retries, authentication, access controls, and appropriate network boundaries.
| Trend | Post Mentions | Total Month Mentions | Posts | Companies | MoM |
|---|---|---|---|---|---|
| MCP | 62 | 2,241 | 148 | 72 | -74% |
| LLM | 3 | 747 | 162 | 79 | -85% |
| AI Agents | 1 | 931 | 231 | 103 | -84% |
| Kubernetes | 1 | 956 | 75 | 30 | -73% |
| OpenTelemetry | 1 | 125 | 18 | 15 | -83% |
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.