MCP and TIA Portal: AI with tool access

A chat window alone does not make an AI assistant for engineering. It gets interesting when the model is allowed to call tools: read the project tree, fetch block source code, generate and import SCL, compile. The open standard for this is the Model Context Protocol (MCP).

What MCP is

MCP connects an AI assistant (the client) to servers that provide tools and data. The assistant sees which tools exist, calls them and gets the results back as context. Instead of copy-paste ping-pong between chat and IDE you get a closed loop: generate code, import it, see the compiler errors, fix them.

What that means in the TIA world

The community is currently building MCP servers that open up TIA Openness to generic assistants. AnyAutomation Studio takes the opposite route: the IDE ships its AI chat with built-in TIA tools (search the project tree, read block sources, create, import and export objects) and is an MCP client on top. You attach external servers (GitHub, Notion, Figma or your own internal services), and their tools line up in the same chat next to the TIA tools. Details in the docs.

Approvals instead of blind flight

Tool access needs limits. Read-only tools may run silently; anything that changes the project asks first. In Studio every tool has an approval level, and destructive actions warn explicitly. And the last line holds regardless: nothing goes to the PLC without you.

Privacy

MCP does not change the model question: you decide which LLM does the thinking, local models included. For confidential plants, project data stays in-house.

Conclusion

MCP turns a chat assistant into an engineering tool: with access to the real project, but behind approvals. If you want to use AI in TIA Portal seriously, you need exactly this integration, not another chat window. For tools built to your company standard there are Custom Extensions.