AI Dev SDK EARLY ACCESS

Last updated:
Edit on GitHub
Warning

Before you use AI-related tools, consult your legal department.

Overview

When you use AI tools to write code, they rely on general patterns, which leads to mistakes and requires you to repeatedly explain your project structure.

The AI Dev SDK makes your AI assistant Spryker-aware. It ships:

  • Skills and agents — packaged workflows for Spryker work: project setup, feature customization, bugfixing, upgrades, testing, and code review. See Workflows, Skills, and Agents.
  • Rules and a project context file — Spryker’s coding conventions and your project’s specifics, loaded into every session.
  • An MCP server (MCP — Model Context Protocol, a standard that lets AI tools query external systems) — live access to your running application: transfer objects, module interfaces, order management system (OMS) state machines, and read-only database queries. See AI Dev MCP Server.

The SDK is delivered by the spryker-sdk/ai-dev module and, for Claude Code, by the spryker-ai-dev-sdk plugin. To set either of them up, see Installation.

How it works

The skills, agents, and rules are files installed into your AI tool’s directories — through the Claude Code plugin (for Claude Code) or the ai-dev:setup console command (for other assistants). The assistant loads the rules and the context file into every session, and when you describe a task, it picks the matching skill and follows a tested Spryker workflow instead of improvising.

The MCP server covers what static files cannot: facts about your running application. It runs as a console command inside your project’s Docker container; when the assistant needs a transfer structure, an OMS transition, or a data check, it queries the server and works with your actual project state instead of assumptions.

The result: the generated code fits your project from the start, and you spend less time correcting mistakes and explaining Spryker concepts.

Key capabilities

Ready-made workflows

Four orchestrated workflows drive multi-step work end to end, and the profiler workflow turns the WebProfiler into hard numbers. They are all documented under Workflows, Skills, and Agents:

Spryker-aware code review

The spryker-code-reviewer agent checks changes against Spryker’s coding standards and layer architecture before they reach a pull request.

Faster Spryker-specific answers

AI can search Spryker documentation instead of requiring you to explain basic concepts or guessing how features work.

Smarter code generation

AI can look up your actual transfer objects, module dependencies, and interface methods so that the generated code matches your project structure instead of relying on assumptions.

OMS debugging made easy

AI can analyze your OMS flows to find possible next states, transitions, conditions, and timeouts for any order or state. This capability is especially helpful when you work with complex OMS schemas. You no longer need to manually follow arrows in large diagrams.

Working with complex data imports

AI can analyze, modify, and transform multi-column CSV files correctly. This task normally requires significant manual effort.

Database queries

AI can execute read-only database queries to inspect data when debugging issues.

Extensible for your project

You can extend the MCP server with custom plugins to add new tools — see Extension points. The shipped skills are plain Markdown files, so you can also use them as a template for your team’s own workflows.

Next steps