Add blog section to the docs

Signed-off-by: Christian Tzolov <christian.tzolov@broadcom.com>

docs/blog/.authors.yml

@@ -0,0 +1,5 @@
+authors:
+  mcp-team:
+    name: MCP Java SDK Team
+    description: Maintainers of the MCP Java SDK
+    avatar: https://github.com/modelcontextprotocol.png

docs/blog/index.md

@@ -0,0 +1 @@
+# Blog

docs/blog/posts/mcp-server-performance-benchmark.md

@@ -0,0 +1,72 @@
+---
+date: 2026-02-15
+authors:
+  - mcp-team
+categories:
+  - Performance
+  - Benchmarks
+---
+
+# Java Leads MCP Server Performance Benchmarks with Sub-Millisecond Latency
+
+A comprehensive independent benchmark of MCP server implementations across four major languages puts Java at the top of the performance charts — delivering sub-millisecond latency, the highest throughput, and the best CPU efficiency of all tested platforms.
+
+<!-- more -->
+
+## The Benchmark
+
+[TM Dev Lab](https://www.tmdevlab.com/mcp-server-performance-benchmark.html) published a rigorous performance comparison of MCP server implementations spanning **3.9 million total requests** across three independent test rounds. The benchmark evaluated four implementations under identical conditions:
+
+- **Java** — Spring Boot 4.0.0 + Spring AI 2.0.0-M2 on Java 21
+- **Go** — Official MCP SDK v1.2.0
+- **Node.js** — @modelcontextprotocol/sdk v1.26.0
+- **Python** — FastMCP 2.12.0+ with FastAPI 0.109.0+
+
+Each server was tested with 50 concurrent virtual users over 5-minute sustained runs in Docker containers (1-core CPU, 1GB memory) on Ubuntu 24.04.3 LTS. Four standardized benchmark tools measured CPU-intensive, I/O-intensive, data transformation, and latency-handling scenarios — all with a **0% error rate** across every implementation.
+
+## Java's Performance Highlights
+
+The results speak for themselves:
+
+| Server     | Avg Latency | Throughput (RPS) | CPU Efficiency (RPS/CPU%) |
+|------------|-------------|------------------|---------------------------|
+| **Java**   | **0.835 ms** | **1,624**       | **57.2**                  |
+| Go         | 0.855 ms    | 1,624            | 50.4                      |
+| Node.js    | 10.66 ms    | 559              | 5.7                       |
+| Python     | 26.45 ms    | 292              | 3.2                       |
+
+```mermaid
+---
+config:
+    xyChart:
+        width: 700
+        height: 400
+    themeVariables:
+        xyChart:
+            backgroundColor: transparent
+---
+xychart-beta
+    title "Average Latency Comparison (milliseconds)"
+    x-axis [Java, Go, "Node.js", Python]
+    y-axis "Latency (ms)" 0 --> 30
+    bar [0.84, 0.86, 10.66, 26.45]
+```
+
+Java achieved the **lowest average latency** at 0.835 ms — edging out Go's 0.855 ms — while matching its throughput at 1,624 requests per second. Where Java truly stands out is **CPU efficiency**: at 57.2 RPS per CPU%, it extracts more performance per compute cycle than any other implementation, including Go (50.4).
+
+In CPU-bound workloads like Fibonacci calculation, Java excelled with a **0.369 ms** response time, showcasing the JVM's highly optimized just-in-time compilation.
+
+## A Clear Performance Tier
+
+The benchmark reveals two distinct performance tiers:
+
+- **High-performance tier**: Java and Go deliver sub-millisecond latencies and 1,600+ RPS
+- **Standard tier**: Node.js (12x slower) and Python (31x slower) trail significantly
+
+Java's throughput is **2.9x higher than Node.js** and **5.6x higher than Python**. For latency-sensitive MCP deployments, the difference is even more pronounced — Java responds **12.8x faster than Node.js** and **31.7x faster than Python**.
+
+## What This Means for MCP Developers
+
+For teams building production MCP servers that need to handle high concurrency and low-latency tool interactions, Java with Spring Boot and Spring AI provides a battle-tested, high-performance foundation. The JVM's mature ecosystem, strong typing, and proven scalability make it an excellent choice for enterprise MCP deployments where performance and reliability are paramount.
+
+The full benchmark details, methodology, and raw data are available at [TM Dev Lab](https://www.tmdevlab.com/mcp-server-performance-benchmark.html).

mkdocs.yml

@@ -48,12 +48,18 @@ nav:
   - Contributing:
     - Contributing Guide: contribute.md
     - Documentation: development.md
+  - Blog:
+    - blog/index.md
   - API Reference: https://javadoc.io/doc/io.modelcontextprotocol.sdk/mcp-core/latest
 
 markdown_extensions:
   - admonition
   - pymdownx.details
-  - pymdownx.superfences
+  - pymdownx.superfences:
+      custom_fences:
+        - name: mermaid
+          class: mermaid
+          format: !!python/name:pymdownx.superfences.fence_code_format
   - pymdownx.tabbed:
       alternate_style: true
   - pymdownx.highlight:
@@ -77,6 +83,7 @@ markdown_extensions:
 
 plugins:
   - search
+  - blog
 
 extra:
   social: