MATLAB AI Agent SDK

MATLAB® AI Agent SDK lets you build and run AI agents in MATLAB.

• Create agents based on OpenAI® Chat Completions API, OpenAI Responses API, Ollama™, or OpenAI-compatible APIs.
• Integrate LLMs and agentic workflows into your workflows in a targeted manner, retaining deterministic workflows when those are more suitable.
• Create agentic tools that use:
  • large amounts of data
  • any MATLAB data type

Research Preview

This SDK is a Research Preview under active development and APIs may change.

Please leave feedback, report bugs and feature requests via Issues. We review all contributions, but we do not merge external pull requests. See CONTRIBUTING.md for details.

Help us improve the MATLAB AI Agent SDK by completing our short feedback survey.

Setup

Using this add-on requires MATLAB R2023a or newer.

You can use the add-on in MATLAB Online™ by clicking this link:

Install by using Add-On Explorer

The recommended way of using the add-on on an installed version of MATLAB is to use the Add-On Explorer.

1. In MATLAB, go to the Home tab, and in the Environment section, click the Add-Ons icon.
2. In the Add-On Explorer, search for "MATLAB AI Agent SDK".
3. Select Install.

Install by cloning GitHub repository

To use the add-on on an installed version of MATLAB, you can clone the GitHub repository. In the MATLAB Command Window, run this command:

>> !git clone https://github.com/matlab/matlab-ai-agent-sdk.git

To run code from the add-on outside of the installation directory, add the path to the installation directory.

>> addpath("path/to/matlab-ai-agent-sdk")

OpenAI

Using the OpenAI API requires an OpenAI API key. For information on how to obtain one, as well as pricing, terms and conditions of use, and available models, see the OpenAI documentation.

Set your key as an environment variable in a .env file:

OPENAI_API_KEY=<your key>

Then load it in MATLAB.

loadenv(".env")

OpenAI-Compatible APIs

To connect to APIs that are compatible with the OpenAI Responses API or Chat Completions API, set your key as an environment variable in a .env file:

OPENAI_API_KEY=<your key>

If your API does not need an API key, set the environment variable OPENAI_API_KEY to "EMPTY".

Then, when creating an LLM client, set the api argument of the aisdk.LLMClient function to "openai" or "openai-responses", depending on which OpenAI API your API is compatible with. Set the BaseURL name-value argument to the base URL of the API. For example, for both OpenAI APIs, BaseURL is "https://api.openai.com/v1".

Ollama

To connect to local or remote Ollama models, first install Ollama.

After you have installed Ollama, you can install models from the MATLAB Command Window:

!ollama pull <modelname>

Get Started

Create an LLM client by using the aisdk.LLMClient function and using the API and the model name as input arguments, for example:

clientOpenAI = aisdk.LLMClient("openai", "gpt-4.1-mini");
clientOllama = aisdk.LLMClient("ollama", "<model-name>");

Then, generate text by using the generate function.

text = generate(client, "This is an example prompt.")

text = 
    "This is an example reponse."

Create Chat With LLM

This example shows how to create a conversation with an LLM and automatically keep track of the message history.

Create the agent from an LLM client client by using the aisdk.AIAgent function. Provide a system prompt.

systemPrompt = "Reply as if you are writing telegrams.";
agent = aisdk.AIAgent(client,SystemPrompt=systemPrompt);

Run the agent by using the run function. Provide a prompt.

prompt = "TOMATO FRUIT OR VEGETABLE STOP";
run(agent,prompt)

ans = 

    "TOMATO TECHNICALLY A FRUIT STOP COMMONLY USED AS VEGETABLE IN CULINARY CONTEXT STOP END OF TRANSMISSION."

Ask a follow up question by using the run function.

run(agent,"HOW ABOUT AVOCADO STOP")

ans = 

    "AVOCADO ALSO A FRUIT STOP KNOWN AS ALLIGATOR PEAR STOP HIGH IN HEALTHY FATS AND NUTRIENTS STOP END OF TRANSMISSION."

Inspect the chat history by using the Messages property of the agent.

agent.Messages

ans = 

  1×4 LLMTextMessage array with messages:

    1    User         Text    "TOMATO FRUIT OR VEGETABLE STOP"
    2    Assistant    Text    "TOMATO TECHNICALLY A FRUIT STOP COMMONLY USED AS VEGETABLE I..."
    3    User         Text    "HOW ABOUT AVOCADO STOP"
    4    Assistant    Text    "AVOCADO ALSO A FRUIT STOP KNOWN AS ALLIGATOR PEAR STOP HIGH ..."

Create AI Agent With Tools

This example shows how to create an AI agent with a set of tools.

Create a function that counts the number of times a letter appears in a word.

function numLetter = countLetters(word,letter)
    numLetter = count(word,letter);
end

Create a tool from the countLetters function by using the aisdk.LLMTool function. Add information about input and output arguments to the tool by using the aisdk.LLMToolArgument function.

tool = aisdk.LLMTool(@countLetters);
tool.InputArguments(1) = aisdk.LLMToolArgument("word",DataType="string");
tool.InputArguments(2) = aisdk.LLMToolArgument("letter",DataType="string");
tool.OutputArguments = aisdk.LLMToolArgument("numLetter",DataType="number");

Create the agent from an LLM client client by using the aisdk.AIAgent function.

agent = aisdk.AIAgent(client,Tools=tool);

Run the agent by using the run function.

run(agent,"How many times is the letter r in the word strawberry?")

ans = "The letter "r" appears 3 times in the word "strawberry.""

Configure Tool to Use Agent Workspace

This example shows how to configure an LLM tool to use data from the agent workspace as input or output data.

The eig function calculates the eigenvectors and eigenvalues of matrices. Vectors and matrices can contain a lot of numerical data. Instead of sending all this data to an LLM, which would cost tokens, keep the data in the agent workspace and configure your tools to work on that workspace.

Create a function called eigTool.

• The first input argument of the function must be a structure array. Call the argument ws.

• The last output argument of the function must be the same structure array.

To allow the agent to understand the outcome of the tool call, add another output argument, observation, that contains a natural language description of the outcome of the tool call. Describe the outcome using the observation output argument.

function [observation,ws] = eigTool(ws)
% Compute the eigenvalues of a matrix
ws.eigenvalues = eig(ws.matrix);
observation = "Eigenvalues added to the workspace as a variable called eigenvalues.";
end

Create an LLM tool from the eigTool function by using the aisdk.LLMTool function. Set the Workspace name-value argument to "agent".

tool = aisdk.LLMTool(@eigTool,Workspace="agent");

You can now add the tool to an agent agentWithWorkspace. Add a matrix A to the agent workspace by setting the Workspace property. Call the field matrix to match the field name in the tool definition.

agentWithWorkspace.Tools = tool;
agentWithWorkspace.Workspace.matrix = A;

Functions

Function	Description
aisdk.AIAgent	Build AI agent
run	Run AI agent
loadSkill	Load skill into agent
resetApproval	Reset approval status of agentic tools
aisdk.LLMClient	Connect to third-party LLM API
OpenAIClient	Client for OpenAI Chat Completions API
OpenAIResponsesClient	Client for OpenAI Responses API
OllamaClient	Client for Ollama API
generate	Generate output from LLMs
aisdk.LLMTool	Tool for AI agent
LocalLLMTool	AI agent tool from MATLAB function
MCPTool	AI agent tool from MCP server
aisdk.MCPClient	Connect to MCP server
select	Select LLM tool from tool array
evaluate	Evaluate LLM tool
aisdk.LLMToolArgument	Argument for LLM tool

Examples

Example	Description
SolveSimpleMathProblemUsingAIAgent.m	Solve a simple math problem by using an AI agent
CreateSimpleChatbot.m	Create a simple, interactive ChatBot in the MATLAB® Command Window
AnalyzeTextDataUsingLLMTool.m	Use tool calling to create a table of customer information from a text data set containing that information in natural language
FitPolynomialToDataUsingAIAgent.m	Build an AI agent to fit polynomials up to ninth degree to data and choose the best fit
OrchestrateMultipleSubagents.m	Create an agent with multiple subagents
CallMCPClientUsingAIAgent.m	Create and run an AI agent that uses tools provided by an MCP server
LabelImagesUsingLLM.m	Categorize and label images by using a large language model

License

The license is available in the LICENSE file in this GitHub repository.

Contact

To ask questions, report issues, or request technical support, open an Issue.

---

Copyright 2026 The MathWorks, Inc.