{"id":14670,"date":"2026-06-17T16:50:11","date_gmt":"2026-06-17T07:50:11","guid":{"rendered":"https:\/\/blogs.mathworks.com\/japan-community\/?p=14670"},"modified":"2026-06-17T16:50:11","modified_gmt":"2026-06-17T07:50:11","slug":"giving-local-ai-agents-the-ability-to-use-matlab-with-mcp-jp","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/japan-community\/2026\/06\/17\/giving-local-ai-agents-the-ability-to-use-matlab-with-mcp-jp\/","title":{"rendered":"MCP \u3067\u5b9f\u73fe\u3059\u308b\u3001\u30ed\u30fc\u30ab\u30eb AI \u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306b\u3088\u308b MATLAB \u6d3b\u7528"},"content":{"rendered":"
\n
\u203b\u3053\u306e\u6295\u7a3f\u306f 2026 \u5e74 1 \u6708 21 \u65e5\u306b Artificial Intelligence \u3078 \u6295\u7a3f<\/a>\u3055\u308c\u305f\u3082\u306e\u306e\u6284\u8a33\u3067\u3059\u3002<\/div>\n
—<\/h6>\n
\n
\n
\n
\n
<\/h6>\n\n\n\n
\"\"<\/td>\nGuest writer: <\/strong>Abhijit Bhattacharjee<\/a><\/p>\n

Abhijit \u306f MathWorks \u306e\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u30a8\u30f3\u30b8\u30cb\u30a2\u3067\u3001AI \u306e\u5c02\u9580\u5bb6\u3067\u3059\u3002\u3053\u3053\u6570\u304b\u6708\u306f\u7279\u306b Agentic AI \u3092\u4e2d\u5fc3\u306b\u3001\u3053\u306e\u5206\u91ce\u306e\u6700\u65b0\u304b\u3064\u6700\u5148\u7aef\u306e\u6280\u8853\u3067\u304a\u5ba2\u69d8\u3092\u652f\u63f4\u3057\u3066\u3044\u307e\u3059\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/div>\n

<\/h6>\n
\n
<\/div>\n<\/div>\n<\/div>\n
<\/div>\n
\u540c\u50da Mike<\/a> \u306e\u3059\u3070\u3089\u3057\u3044\u8a18\u4e8b NVIDIA DGX Spark \u4e0a\u3067 LLM \u3092\u52d5\u304b\u3057\u3001MATLAB \u304b\u3089\u63a5\u7d9a\u3059\u308b\u65b9\u6cd5<\/a> \u3092\u8e0f\u307e\u3048\u3066\u3001\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3092\u9006\u8ee2\u3067\u304d\u306a\u3044\u304b\u8a66\u3057\u3066\u307f\u305f\u304f\u306a\u308a\u307e\u3057\u305f\u3002\u3064\u307e\u308a\u3001MATLAB \u304c LLM\uff08\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\uff09\u3092\u547c\u3073\u51fa\u3059\u4ee3\u308f\u308a\u306b\u3001\u30ed\u30fc\u30ab\u30eb LLM \u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306b MATLAB \u3092\u547c\u3073\u51fa\u3055\u305b\u305f\u3089\u3069\u3046\u3060\u308d\u3046\uff1f<\/span>\u3068\u3044\u3046\u3053\u3068\u3067\u3059\u3002\u305d\u3046\u3059\u308c\u3070 MATLAB \u5168\u4f53\u305d\u306e\u3082\u306e\u304c\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u9053\u5177\u3068\u306a\u308a\u3001LLM \u304c\u66f8\u3044\u305f\u30b3\u30fc\u30c9\u3092\u5b9f\u884c\u3057\u3001\u691c\u8a3c\u3057\u3001\u51e6\u7406\u3059\u308b\u6e96\u5099\u304c\u6574\u3044\u307e\u3059\u3002<\/div>\n
\u3057\u304b\u3082\u4e00\u756a\u826f\u3044\u306e\u306f\u4f55\u304b\u3068\u3044\u3046\u3068\u3001\u4f55\u3082\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u306e\u5916\u306b\u51fa\u308b\u3053\u3068\u306a\u304f\u3001\u3059\u3079\u3066\u30ed\u30fc\u30ab\u30eb\u3067\u51e6\u7406\u3055\u308c\u308b\u3053\u3068\u3067\u3059\u3002<\/div>\n
\u5e78\u3044\u306b\u3082\u79c1\u306f\u624b\u5143\u306b NVIDIA DGX Spark \u5b9f\u6a5f\u3092\u5165\u624b\u3057\u307e\u3057\u305f\u3002<\/div>\n
\"\"<\/div>\n
\u3053\u3061\u3089<\/a>\u3067 DGX Spark \u306e\u4ed5\u69d8\u3092\u3055\u3089\u306b\u78ba\u8a8d\u3067\u304d\u307e\u3059\u304c\u3001\u3053\u3053\u3067\u91cd\u8981\u306a\u306e\u306f\u3001\u672c\u683c\u7684\u306a\u30a8\u30f3\u30b8\u30cb\u30a2\u30ea\u30f3\u30b0\u4f5c\u696d\u304c\u3067\u304d\u308b\u898f\u6a21\u306e\u30ed\u30fc\u30ab\u30eb LLM \u3092\u6271\u3048\u308b\u3060\u3051\u306e\u5341\u5206\u306a VRAM\uff08128GB\uff09\u3092\u5099\u3048\u3066\u3044\u308b\u3053\u3068\u3067\u3059\u3002<\/div>\n

\u30ed\u30fc\u30ab\u30eb LLM \u306e\u30bb\u30c3\u30c8\u30a2\u30c3\u30d7<\/h2>\n
Mike \u3068\u540c\u3058\u304f\u3001\u79c1\u306f gpt-oss-120b<\/a> \u30e2\u30c7\u30eb\u3092\u9078\u3073\u307e\u3057\u305f\u3002\u4fe1\u983c\u3067\u304d\u308b\u30c4\u30fc\u30eb\u547c\u3073\u51fa\u3057\u6027\u80fd\u3068\u3001mixture-of-experts\uff08MoE\uff09<\/a> \u3068\u3057\u3066\u306e\u9ad8\u3044\u6027\u80fd\u304c\u7406\u7531\u3067\u3001\u3053\u308c\u306f 1200 \u5104\u500b\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u3046\u3061\u4e00\u90e8\u3060\u3051\u3092\u80fd\u52d5\u7684\u306b\u4f7f\u3046\u305f\u3081\u3001\u8a08\u7b97\u8ca0\u8377\u3092\u6291\u3048\u3089\u308c\u307e\u3059\u3002\u305f\u3060\u3057\u79c1\u306f\u5f7c\u3088\u308a\u5c11\u3057\u8907\u96d1\u306a\u9053\u3092\u9078\u3073\u3001\u30e2\u30c7\u30eb\u5b9f\u884c\u306e\u305f\u3081\u306b llama.cpp<\/a> \u3092\u30bd\u30fc\u30b9\u304b\u3089\u30b3\u30f3\u30d1\u30a4\u30eb\u3059\u308b\u3053\u3068\u306b\u3057\u307e\u3057\u305f\u3002\u3053\u308c\u3082\u30d0\u30c3\u30af\u30a8\u30f3\u30c9\u306b llama.cpp \u3092\u4f7f\u3046 Ollama \u3088\u308a\u3001\u3055\u3089\u306b\u591a\u304f\u306e tokens-per-second\uff08tps\uff09\u3092\u5f15\u304d\u51fa\u305b\u306a\u3044\u304b\u8a66\u3057\u305f\u304b\u3063\u305f\u304b\u3089\u3067\u3059\u3002\u96e3\u3057\u305d\u3046\u306b\u805e\u3053\u3048\u308b\u304b\u3082\u3057\u308c\u307e\u305b\u3093\u304c\u3001\u30ea\u30dd\u30b8\u30c8\u30ea\u306b\u3042\u308b\u5305\u62ec\u7684\u306a\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u306e\u304a\u304b\u3052\u3067\u3001\u5b9f\u969b\u306b\u306f\u305d\u308c\u307b\u3069\u5927\u5909\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u3002<\/div>\n
\u30ed\u30fc\u30ab\u30eb\u30e2\u30c7\u30eb\u3092\u4f7f\u3046\u306b\u306f\u3001\u9069\u5207\u306a\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u4ed8\u3051\u3066 llama-server \u3092\u8d77\u52d5\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002\u4ee5\u4e0b\u306f\u3001DGX Spark \u4e0a\u3067 gpt-oss-120b \u3092\u52d5\u304b\u3059\u305f\u3081\u306b\u6700\u9069\u5316\u3057\u305f\u79c1\u306e\u8d77\u52d5\u30b3\u30de\u30f3\u30c9\u3067\u3059\u3002<\/div>\n
llama-server \\<\/span><\/div>\n
-m models\/gpt-oss-120b\/Q4_K_M\/gpt-oss-120b-Q4_K_M-00001-of-00002.gguf \\<\/span><\/div>\n
–host 0.0.0.0 –port 8080 \\<\/span><\/div>\n
-ngl 999 \\<\/span><\/div>\n
-fa on \\<\/span><\/div>\n
-t 10 \\<\/span><\/div>\n
-c 0 \\<\/span><\/div>\n
-b 2048 -ub 2048 \\<\/span><\/div>\n
-ctk q8_0 -ctv q8_0 \\<\/span><\/div>\n
–no-mmap \\<\/span><\/div>\n
–jinja \\<\/span><\/div>\n
–reasoning-format auto \\<\/span><\/div>\n
–chat-template-kwargs “{\\”reasoning_effort\\”: \\”medium\\”}” \\<\/span><\/div>\n
–temp 1.0 \\<\/span><\/div>\n
–top-p 1.0 \\<\/span><\/div>\n
–top-k 0<\/span><\/div>\n
\u3053\u306e\u3088\u3046\u306b\u5b9f\u884c\u3059\u308b\u3053\u3068\u3067\u3001\u3044\u304f\u3064\u304b\u306e\u6700\u9069\u5316\u3092\u6307\u5b9a\u3067\u304d\u307e\u3057\u305f\u3002\u305f\u3068\u3048\u3070 –no-mmap<\/span> \u306f\u30e1\u30e2\u30ea\u30de\u30c3\u30d7\u30c9\u30d5\u30a1\u30a4\u30eb I\/O \u3092\u7121\u52b9\u306b\u3057\u307e\u3059\uff08DGX Spark \u306e\u3088\u3046\u306a\u30e6\u30cb\u30d5\u30a1\u30a4\u30c9\u30e1\u30e2\u30ea\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3067\u306f\u4e0d\u8981\u3067\u3059\uff09\u3002\u307e\u305f -fa on<\/span> \u306f flash attention \u30ab\u30fc\u30cd\u30eb\u3092\u6709\u52b9\u306b\u3057\u3001\u9577\u3044\u30c8\u30fc\u30af\u30f3\u5217\u306e\u51e6\u7406\u3092\u9ad8\u901f\u5316\u3057\u307e\u3059\u3002<\/div>\n

\u30ed\u30fc\u30ab\u30eb\u30e2\u30c7\u30eb\u3067\u884c\u3046 Agentic AI \u30b3\u30fc\u30c7\u30a3\u30f3\u30b0<\/h3>\n
\u3055\u3066\u3001MATLAB \u3092\u547c\u3073\u51fa\u3057\u53ef\u80fd\u306a\u30c4\u30fc\u30eb\u3068\u3057\u3066\u4f7f\u3046\u524d\u306b\u3001\u307e\u305a\u306f MATLAB \u306e\u5916\u304b\u3089\u3053\u306e\u30ed\u30fc\u30ab\u30eb\u30e2\u30c7\u30eb\u3068\u3084\u308a\u53d6\u308a\u3059\u308b\u65b9\u6cd5\u304c\u5fc5\u8981\u3067\u3057\u305f\u3002\u3053\u3053\u308c\u306f\u4eba\u6c17\u306e Agentic AI \u7528\u30bf\u30fc\u30df\u30ca\u30eb\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3 OpenCode<\/a> \u3092\u30cf\u30fc\u30cd\u30b9\u3068\u3057\u3066\u5229\u7528\u3057\u307e\u3059\u3002OpenCode \u3067\u306f\u30ed\u30fc\u30ab\u30eb\u30e2\u30c7\u30eb\u3092\u3055\u307e\u3056\u307e\u306a\u65b9\u6cd5\u3067\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002\u305d\u306e\u7c21\u5358\u306a\u65b9\u6cd5\u306e\u4e00\u3064\u304c\u3001opencode.json \u8a2d\u5b9a\u30d5\u30a1\u30a4\u30eb\u3092\u4f7f\u3046\u3053\u3068\u3067\u3059\u3002<\/div>\n
\u3082\u3046\u4e00\u3064\u3084\u308a\u305f\u3044\u3053\u3068\u306f\u3001OpenCode \u304c\u79c1\u306e\u30de\u30b7\u30f3\u4e0a\u306e\u30ed\u30fc\u30ab\u30eb MATLAB MCP Core Server<\/a> \u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3092\u63a5\u7d9a\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u3067\u3059\u3002<\/span>Model Context Protocol\uff08MCP\uff09\u3068\u305d\u306e\u4ed5\u7d44\u307f\u3092\u77e5\u308a\u305f\u3044\u5834\u5408\u306f\u3001\u3053\u306e\u8a71\u984c\u306b\u95a2\u3059\u308b\u6700\u8fd1\u306e\u30d6\u30ed\u30b0\u8a18\u4e8b\u3001\u305f\u3068\u3048\u3070 \u3053\u3061\u3089<\/a> \u3092\u305c\u3072\u3054\u89a7\u304f\u3060\u3055\u3044\u3002<\/div>\n
\u3082\u3046\u5c11\u3057 JSON \u3092\u8abf\u6574\u3057\u3001\u6700\u7d42\u7684\u306b \u30ed\u30fc\u30ab\u30eb LLM\uff08\u30a8\u30f3\u30c9\u30dd\u30a4\u30f3\u30c8\u30a2\u30c9\u30ec\u30b9\u306f 192.168.108.170<\/span>\uff09\u3068\u30ed\u30fc\u30ab\u30eb MATLAB MCP \u63a5\u7d9a\u306e\u4e21\u65b9\u3092\u8a2d\u5b9a\u3059\u308b\u5b8c\u5168\u306a opencode.json \u30d5\u30a1\u30a4\u30eb\u306f\u6b21\u306e\u3088\u3046\u306b\u306a\u308a\u307e\u3057\u305f\u3002<\/div>\n
{<\/span><\/div>\n
“$schema”<\/span>: <\/span>“https:\/\/opencode.ai\/config.json”<\/span>,<\/span><\/div>\n
“provider”<\/span>: {<\/span><\/div>\n
“llama.cpp”<\/span>: {<\/span><\/div>\n
“npm”<\/span>: <\/span>“@ai-sdk\/openai-compatible”<\/span>,<\/span><\/div>\n
“name”<\/span>: <\/span>“llama.cpp (abhijit-spark)”<\/span>,<\/span><\/div>\n
“options”<\/span>: {<\/span><\/div>\n
“baseURL”<\/span>: <\/span>“http:\/\/192.168.108.170:8080\/v1”<\/span><\/div>\n
},<\/span><\/div>\n
“models”<\/span>: {<\/span><\/div>\n
“gpt-oss-120b-GGUF”<\/span>: {<\/span><\/div>\n
“name”<\/span>: <\/span>“gpt-oss-120b”<\/span><\/div>\n
}<\/span><\/div>\n
}<\/span><\/div>\n
}<\/span><\/div>\n
},<\/span><\/div>\n
“mcp”<\/span>: {<\/span><\/div>\n
“MATLAB MCP”<\/span>: {<\/span><\/div>\n
“type”<\/span>: <\/span>“local”<\/span>,<\/span><\/div>\n
“command”<\/span>: [<\/span><\/div>\n
“C:<\/span>\\\\<\/span>Users<\/span>\\\\<\/span>abhijit<\/span>\\\\<\/span>Local Content<\/span>\\\\<\/span>Apps<\/span>\\\\<\/span>matlab-mcp-core-server-win64.exe”<\/span>, <\/span><\/div>\n
“–matlab-root=C:<\/span>\\\\<\/span>Program Files<\/span>\\\\<\/span>MATLAB<\/span>\\\\<\/span>R2025b”<\/span><\/div>\n
],<\/span><\/div>\n
“enabled”<\/span>: <\/span>true<\/span>,<\/span><\/div>\n
}<\/span><\/div>\n
}<\/span><\/div>\n
}<\/span><\/div>\n
\u4e0a\u8a18\u306e\u3088\u3046\u306b\u3059\u3079\u3066\u8a2d\u5b9a\u3059\u308b\u3068\u3001OpenCode \u3092\u8d77\u52d5\u3057\u305f\u3068\u304d\u306b\u3001\u30ed\u30fc\u30ab\u30eb\u30e2\u30c7\u30eb\u304c\u9078\u629e\u3055\u308c\u5229\u7528\u53ef\u80fd\u3067\u3042\u308b\u3053\u3068\u3001\u305d\u3057\u3066 MATLAB MCP \u304c\u63a5\u7d9a\u3055\u308c\u52d5\u4f5c\u3057\u3066\u3044\u308b\u3053\u3068\u3092\u793a\u3059\u753b\u9762\u304c\u8868\u793a\u3055\u308c\u307e\u3059\u3002<\/div>\n
\"\"<\/div>\n

\u30d0\u30a4\u30d6\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u3092\u59cb\u3081\u307e\u3057\u3087\u3046\uff01<\/h2>\n
MATLAB MCP Core Server \u3092\u6709\u52b9\u306b\u3057\u305f\u72b6\u614b\u3067\u3053\u308c\u304c\u3069\u3046\u52d5\u304f\u304b\u3092\u78ba\u304b\u3081\u308b\u305f\u3081\u306b\u3001MATLAB \u306b\u6a19\u6e96\u3067\u4ed8\u5c5e\u3059\u308b\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u4f7f\u3063\u305f\u30c7\u30fc\u30bf\u89e3\u6790\u3092\u884c\u3046\u6b21\u306e\u30d7\u30ed\u30f3\u30d7\u30c8\u3092\u5165\u529b\u3057\u307e\u3059\u3002<\/div>\n
Create a MATLAB script that loads carbig dataset and creates 3 plots showcasing how different car properties affect MPG. Use default positions for any figures you create.\uff08carbig \u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u8aad\u307f\u8fbc\u307f\u3001\u3055\u307e\u3056\u307e\u306a\u8eca\u4e21\u7279\u6027\u304c MPG \u306b\u3069\u306e\u3088\u3046\u306b\u5f71\u97ff\u3059\u308b\u304b\u3092\u793a\u3059 3 \u3064\u306e\u30d7\u30ed\u30c3\u30c8\u3092\u4f5c\u6210\u3059\u308b MATLAB \u30b9\u30af\u30ea\u30d7\u30c8\u3092\u4f5c\u6210\u3057\u3066\u304f\u3060\u3055\u3044\u3002\u4f5c\u6210\u3059\u308b\u56f3\u306e\u4f4d\u7f6e\u306f\u3059\u3079\u3066\u65e2\u5b9a\u5024\u3092\u4f7f\u7528\u3057\u3066\u304f\u3060\u3055\u3044\u3002)<\/span><\/div>\n
Train an ensemble regression model using this data to predict MPG from weight, horsepower, displacement, and acceleration. Show me the model statistics and diagnostic plots. (\u3053\u306e\u30c7\u30fc\u30bf\u3092\u4f7f\u3063\u3066\u30a2\u30f3\u30b5\u30f3\u30d6\u30eb\u56de\u5e30\u30e2\u30c7\u30eb\u3092\u5b66\u7fd2\u3057\u3001\u91cd\u91cf\u3001\u99ac\u529b\u3001\u6392\u6c17\u91cf\u3001\u52a0\u901f\u5ea6\u304b\u3089 MPG \u3092\u4e88\u6e2c\u3057\u3066\u304f\u3060\u3055\u3044\u3002\u30e2\u30c7\u30eb\u7d71\u8a08\u91cf\u3068\u8a3a\u65ad\u30d7\u30ed\u30c3\u30c8\u3092\u8868\u793a\u3057\u3066\u304f\u3060\u3055\u3044\u3002)<\/span><\/div>\n
Use the MATLAB MCP to execute and test your code. (MATLAB MCP \u3092\u4f7f\u3063\u3066\u30b3\u30fc\u30c9\u3092\u5b9f\u884c\u3057\u3001\u30c6\u30b9\u30c8\u3057\u3066\u304f\u3060\u3055\u3044\u3002)<\/span><\/div>\n
\u901a\u5e38 Claude Opus \u3084 GPT-5.2 \u306e\u3088\u3046\u306a\u6700\u5148\u7aef\u30e2\u30c7\u30eb\u3067\u306f\u3001MCP \u3092\u547c\u3073\u51fa\u3059\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u3001\u308f\u3056\u308f\u3056\u6307\u5b9a\u3059\u308b\u5fc5\u8981\u306f\u3042\u308a\u307e\u305b\u3093\u3002<\/span>\u305d\u3046\u3057\u305f LLM \u306f\u81ea\u52d5\u7684\u306b\u305d\u308c\u3092\u5224\u65ad\u3057\u3066\u304f\u308c\u307e\u3059\u3002\u305f\u3060\u3001gpt-oss-120b \u306b\u3064\u3044\u3066\u306f\u3001\u3042\u307e\u308a\u9577\u304f\u7a7a\u56de\u308a\u3057\u306a\u3044\u3088\u3046\u3001\u5c11\u3057\u8ffd\u52a0\u306e\u30ac\u30a4\u30c0\u30f3\u30b9\u3092\u4e0e\u3048\u308b\u3053\u3068\u306b\u3057\u307e\u3057\u305f\u3002\u4ee5\u4e0b\u306f\u3053\u306e\u30bb\u30c3\u30b7\u30e7\u30f3\u306e\u52d5\u753b\u3067\u3059\u3002<\/div>\n