SATK v2026.07.02

Quick Start

Fresh Install

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to install, then run:

>> setupAgenticToolkit("install")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

Update Existing Installation

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to update the installer files, then run:

>> setupAgenticToolkit("update")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

What's New

• Skill group rename: Verification, Validation & Test — The agentic-review skill group has been renamed to verification-validation-and-test and is now registered alongside the code-generation group.

• Single-precision conversion skill — New skill in the Fixed-Point Designer group that converts a double-precision Simulink system or subsystem to single precision. Requires Fixed-Point Designer.

• Incremental layout improvements — The autolayout engine now better handles partially-connected blocks when adding new content to a model. Block sizing and connection routing improved.

Bug Fixes

• setupAgenticToolkit errors when using Source parameter — When using local source mode, the installer incorrectly skipped the MCP server binary download. The Source parameter now only controls where toolkit files come from, not whether network downloads are allowed for required components. (#39)

• Legacy MCP toolbox auto-replacement during setup — Setup now automatically detects and replaces the pre-rename MCP toolbox, resolving an issue where MCP tools could fail silently after a binary update.

Full Changelog: SATK-v2026.06.26...SATK-v2026.07.02

SATK v2026.06.26

Quick Start

Fresh Install

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to install, then run:

>> setupAgenticToolkit("install")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

Update Existing Installation

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to update the installer files, then run:

>> setupAgenticToolkit("update")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

What's New

• New agentic-review skill group — Two new skills in this group:

  • Check authoring skill — Guides agents through creating custom edit-time and compile-time checks using the Simulink Check API. Requires Simulink Check.
  • Model compliance checking skill — Run and interpret Model Advisor and Simulink Check compliance results. Agents can analyze compliance findings and suggest remediation.

• New model_scan util — New util that efficiently reads Simulink models (.slx and .mdl). Used within model_edit.

• Updated documentation — READMEs streamlined for clarity, with newly documented skills and a standalone migration guide.

Bug Fixes

• Knowledge Index check wording improvements (curating-library-kg skill) — Clarified messaging around Knowledge Index validation status so users better understand what's being checked and what actions to take.

Full Changelog: SATK-v2026.06.24...SATK-v2026.06.26

SATK v2026.06.24

Quick Start

Fresh Install

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to install, then run:

>> setupAgenticToolkit("install")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

Update Existing Installation

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to update the installer files, then run:

>> setupAgenticToolkit("update")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

What's New

• Offline/air-gapped install support — When all toolkits are provided via Source= and the MCP server binary already exists on disk, the installer skips all network calls. Enterprise customers on restricted networks can now install without GitHub access.

• Custom library reuse enhancements — Improved change detection for custom library knowledge graphs: only libraries that actually changed are regenerated, significantly reducing rebuild times for multi-library projects.

• Agent MCP config refresh on update — setupAgenticToolkit("update") now automatically rewrites all tracked agent MCP configs when the binary path changes. Previously, agents (Claude Code, Copilot, Codex, Gemini CLI, etc.) could fail to connect to the MCP server after an update because their configs still referenced the old binary name.

• Autolayout improvements — Improvements in Stateflow layout improved routing.

Bug Fixes

• model_check errors with "argument tokenBudget not provided" — Fixed model_check failing when called without the optional tokenBudget parameter.

• create_subsystem ignores port names with string array format — Port names specified as string arrays (e.g., ["data","ctrl"]) were silently dropped. Fixed to handle both cell and string array inputs.

• GettingStarted.mlx sections couldn't run in Live Editor — setupAgenticToolkit commands require command-line interactivity; converted from Live Editor code sections to code examples.

• Simulink toolkit gated on R2023a+ — Setup now checks your MATLAB release before offering the Simulink toolkit. On pre-R2023a releases, setup explains the requirement clearly and auto-selects the MATLAB Agentic Toolkit only. Prevents the confusing "Unable to open P-file" error.

Full Changelog: SATK-v2026.06.12...SATK-v2026.06.24

SATK v2026.06.12

Quick Start

Fresh Install

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to install, then run:

>> setupAgenticToolkit("install")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

Update Existing Installation

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to update the installer files, then run:

>> setupAgenticToolkit("update")

3. Each MATLAB session, initialize the toolkit:

>> addpath("~/.matlab/agentic-toolkits/simulink")
>> satk_initialize

For detailed instructions (manual setup, updating, troubleshooting), see GETTING_STARTED.md.

What's New

• Custom library support for building models — model_edit now resolves blocks from custom-defined Simulink libraries. When a library is configured for a project, the agent prefers library blocks over built-in equivalents. Two new skills support this workflow:

  • configuring-block-policy — Define which blocks agents must, should, or must-not use during model editing.
  • curating-library-kg — Populate and maintain a searchable knowledge graph of your library's blocks.

  The building-simulink-models skill now includes a library setup reference guide.

• Simplified MCP toolbox installation — Setup no longer shells out to a separate process to install the MATLAB MCP connectivity toolbox. The toolbox is now installed in-process, resolving hangs some users experienced during setup.

• MCP server binary rename preparation — Prepares for the upcoming rename of the MCP server binary (effective 6/18). Fresh installs already use the new name. Existing users who run setupAgenticToolkit("update") after 6/18 will transition automatically — no manual action needed.

• Autolayout improvements — Improved automatic layout results for models with highly branched signals.

Bug Fixes

Autolayout (Stateflow)

• Fixed parallel objects not being sorted correctly in some cases, causing incorrect execution order
• Fixed self-loop transition exceptions
• Fixed super-transition label positioning

SATK v2026.06.05

Quick Start

Fresh Install

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to install, then run:

   >> setupAgenticToolkit("install")

3. Each MATLAB session, initialize the toolkit:

   >> addpath("~/.matlab/agentic-toolkits/simulink")
   >> satk_initialize

Update Existing Installation

1. Download agenticToolkitInstaller.mltbx from the release assets below
2. In MATLAB, double-click the .mltbx to update the installer files, then run:

   >> setupAgenticToolkit("update")

3. Each MATLAB session, initialize the toolkit:

   >> addpath("~/.matlab/agentic-toolkits/simulink")
   >> satk_initialize

For detailed instructions (manual setup, updating, troubleshooting), see GETTING_STARTED.md.

What's New

• Native Stateflow editing in model_edit — Stateflow elements can now be created and edited directly via model_edit. Includes autolayout support for chart internals and an updated building-simulink-models skill with Stateflow reference.
• New skill group: Model-Based Systems Engineering — New skills for System Composer:
  • building-architecture-models — Guides creation of architecture models (functional, logical, physical) with System Composer API reference and requirements traceability patterns.
• New managing-simulink-projects skill — Guides MATLAB/Simulink project management: file organization, model references, data dictionaries, source control, and derived file caching.
• filing-bug-reports skill: skill-loading state capture — Bug reports now always record which skills are registered, invoked, name-only, or missing, making it easier to diagnose skill-related failures.
• Setup installer Getting Started guide — New Getting Started Live Script shown during interactive toolbox installation covering basic setup workflows. Improved prompt wording throughout.

Bug Fixes

model_query_params

• Fixed sporadic issue when querying Stateflow elements
• Fixed lookup failures for Stateflow transitions

model_check

• Fixed stale results due to race condition with model_edit changes

model_edit

• Improved error messages when referencing non-existent blocks
• Fixed autolayout failure when operating on subsystems in certain cases

Setup

• Fixed Claude Code deferred tool schema loading (InputValidationError) by adding alwaysLoad: true to MCP config (requires reconfiguring your agent to take effect)
• Fixed Windows legacy binary cleanup failing on locked files and NTFS junctions
• Fixed update check failing for prefixed release tags (SATK-v2026.x.y)
• Fixed Windows backslash paths breaking Codex TOML configuration

SATK v2026.06.01

What's New

• model_check tool — Tool for structural validation of models. Should improve performance of model creation/editing by reducing dangling blocks/unconnected signals.
• Skill group selection — Users can now select which skill groups to enable during setup, reducing context window usage as the skill catalog grows. Supports both interactive and programmatic interfaces.
• model_query_params Stateflow support
• Stateflow autolayout improvements — Optimized transition label positioning and execution order.

Bug Fixes

model_edit

• Fixed issues when configuring certain block types with multiple dialog parameters

model_query_params

• Stateflow and nested-block targets no longer rejected as invalid format

Setup

• Resolved GitHub API rate limit failures during install/update on shared networks
• Fixed --matlab-root incorrectly emitted for existing session mode
• Fixed Windows binary update failing when an agent session is already running

What's New — 2026.05.21

• model_read MATLAB Function block enhancements — Improved model_read performance with MATLAB Function blocks

• Simscape routing improvements — Improved block routing in Simscape diagrams

• Granular container classification — Improved model_overview and model_read representation for Subsystems

• model_check unconnected ports — Structural validation now detects unconnected ports and dangling lines.

Bug Fixes

model_edit

• Improved Fuzzy type matcher for block placement
• Singleton blocks (Ground, Terminator) are reused instead of erroring on duplicate add
• Triggered subsystem event-to-data connections no longer auto-branch to trigger port

model_read

• Product block division expressions now computed correctly (#23)

model_overview

• No longer fails on unsaved in-memory models

Parameters & layout

• Invalid enum values now properly error instead of silently corrupting the model (#20)
• Improvements to layout performance

Setup & initialization

• Resolved issues calling satk_initialize from startup.m (#22)
• Configure no longer overwrites unrelated MCP server entries
• Graceful fallback when Stateflow/Simscape licenses are unavailable or on pre-R2025a

2026.05.07 — Setup overhaul bug fixes

Reminder: starting from 2026.05.06 This is a breaking release. The setup workflow has fundamentally changed. If you previously installed using the agent-driven setup ("Set up the Simulink Agentic Toolkit"), you must uninstall the old installation before upgrading. See migration steps below. This affects all previous installations of MATLAB and Simulink Agentic toolkit and MATLAB MCP.

Bug Fixes

• Restored .claude-plugin/marketplace.json (#14)
• Setup script will no longer delete unrelated MCP servers from config (#15)
• Updated Getting Started info
• Other minor setup bug fixes

See yesterday's release notes below:

🚨⚠️Breaking Change: New Setup Workflow⚠️🚨

This release replaces the agent-driven setup workflow with a new MATLAB function: setupAgenticToolkit. The old workflow is no longer supported.

⚠️⚠️Required: Clean Up Your Previous Installation⚠️⚠️

If you previously set up the toolkit using the agent-driven workflow, you must uninstall the old setup before using the new one. Leftover files and config entries will cause conflicts.

Step 1: Remove old files

If you installed the MCP Add-On toolbox in MATLAB, also uninstall it:

matlab.addons.uninstall("MATLABMCPCoreServerToolkit")

% or 

matlab.addons.uninstall("MATLABMCPCoreServerToolbox")

Step 2: Remove old MCP server entries from agent config files

Remove any matlab or simulink MCP server entries that point to the old locations:

The new setup will write fresh, correct entries during the configure step.

Step 3: (Optional) Delete your old toolkit clone

If you cloned this repository solely for the old agent-driven setup, you can delete it. The new setup downloads toolkit files automatically.

New Setup: setupAgenticToolkit

Download agenticToolkitInstaller.mltbx from the release assets below and double-click it in MATLAB to install via the Add-On Manager. Or install programmatically:

matlab.addons.toolbox.installToolbox("agenticToolkitInstaller.mltbx")

Then run:

setupAgenticToolkit("install")

This single command downloads the MCP server binary, downloads toolkit files, configures your AI coding agent, and installs the MATLAB toolbox for MCP connectivity.

After install, run satk_initialize in MATLAB to load toolkit paths, then restart your agent session.

Full Changelog: 2026.05.06...2026.05.07

2026.05.06 — Setup overhaul

This is a breaking release. The setup workflow has fundamentally changed. If you previously installed using the agent-driven setup ("Set up the Simulink Agentic Toolkit"), you must uninstall the old installation before upgrading. See migration steps below. This affects all previous installations of MATLAB and Simulink Agentic toolkit and MATLAB MCP.

🚨⚠️Breaking Change: New Setup Workflow⚠️🚨

This release replaces the agent-driven setup workflow with a new MATLAB function: setupAgenticToolkit. The old workflow is no longer supported.

⚠️⚠️Required: Clean Up Your Previous Installation⚠️⚠️

If you previously set up the toolkit using the agent-driven workflow, you must uninstall the old setup before using the new one. Leftover files and config entries will cause conflicts.

Step 1: Remove old files

If you installed the MCP Add-On toolbox in MATLAB, also uninstall it:

matlab.addons.uninstall("MATLABMCPCoreServerToolkit")

% or 

matlab.addons.uninstall("MATLABMCPCoreServerToolbox")

Step 2: Remove old MCP server entries from agent config files

Remove any matlab or simulink MCP server entries that point to the old locations:

The new setup will write fresh, correct entries during the configure step.

Step 3: (Optional) Delete your old toolkit clone

If you cloned this repository solely for the old agent-driven setup, you can delete it. The new setup downloads toolkit files automatically.

New Setup: setupAgenticToolkit

Download agenticToolkitInstaller.mltbx from the release assets below and double-click it in MATLAB to install via the Add-On Manager. Or install programmatically:

matlab.addons.toolbox.installToolbox("agenticToolkitInstaller.mltbx")

Then run:

setupAgenticToolkit("install")

This single command downloads the MCP server binary, downloads toolkit files, configures your AI coding agent, and installs the MATLAB toolbox for MCP connectivity.

After install, run satk_initialize in MATLAB to load toolkit paths, then restart your agent session.

What's New

Setup process overhaul — See information above
Improved testing-simulink-models skill — Added more clear instructions for activating the testing-simulink-models skill

Bug Fixes

• Minor Stateflow autolayout fixes.
• Improvements to model_test performance
• MCP tool annotations for all tools (#9)

Other Changes

Setup ships as mltbx — The setup script is now packaged as a MATLAB Add-On (.mltbx) attached to each release for easy installation via the Add-On Manager.

Full Changelog: 2026.05.01...2026.05.06

What's New — 2026.05.01

• Library block support in model_read — Identifies library-linked blocks and labels control ports by function (enable, trigger, reset, ifaction).

• Improved model_edit API — Consistent parameter naming and support for adding library-linked blocks directly.

• Better error reporting in model_test — Errors are now distinguished from failures, with actionable messages for draft-mode type mismatches (#8).

• Skill-aware test authoring — model_test tool description now guides agents to load the testing skill before writing Gherkin files.

Bug Fixes

• Minor improvements for initialization.
• Minor bug fixes in model_edit.
• Minor bug fixes in model_read.

Other Changes

• Simplified version tags — Release tags are now date-based (e.g. 2026.05.01), dropping the previous SATK-v prefix.
• Source available in release assets — Starting with this release, the source .zip attached to each GitHub release is where source files can be retrieved.