AI Integration for TIA Portal (MCP & LLM)
An AI assistant in TIA Portal supports PLC programming with project-wide context: generate SCL blocks, explain and refactor code, and reach the TIA project through MCP tools. AnyAutomation Studio integrates several LLM providers for this.
What is MCP?
The Model Context Protocol (MCP) is an open standard that lets an AI assistant reach tools and data that live outside the model. AnyAutomation Studio supports MCP as a client: you add external MCP servers — Figma, Notion, Linear, Atlassian, a GitHub MCP server, or your own internal services — and their tools become available to the built-in AI chat alongside the TIA-aware tools that ship with Studio.
AnyAutomation Studio does not run as an MCP server that other AI clients connect to, and there is no launch argument that turns it into one. The flow is the other way around: Studio is the client, and you point it at the servers you want to use.
Built-in TIA tools
Studio's AI chat comes with a set of TIA Portal-aware tools that operate against the project you currently have open and connected. The assistant uses them to browse the project tree, read block source, search the hardware catalog, run analyses, and — with your confirmation — create, import, generate, export, or delete project objects. Read tools run silently and feed their results back to the model as context; tools that change the project ask for confirmation in the chat first, and destructive actions state that they cannot be undone. After a confirmed change succeeds, the TIA Portal Project Explorer refreshes automatically so the result is visible.
The TIA tools (except the connect tool) require an active TIA Portal connection. If none is attached, the assistant can attach to a running instance and open a project by path — the same path the Project Explorer uses.
Adding an external MCP server
External MCP servers are managed from the MCP Servers section of the Agent Customizations editor (open it from the chat gear → Open Customizations, or from the Command Palette). To register a server:
- Open the MCP Servers section and choose Add Server.
- Pick a transport — a local subprocess (stdio) or a remote URL (HTTP).
- For a local server, provide the command, arguments, working directory, and any environment variables. For a remote server, provide the endpoint URL and choose an authentication mode.
- Save the server. It appears in the list with a per-server enable toggle and Start / Stop controls.
Once enabled and started, the server's tools join the chat tool picker next to the built-in TIA tools, and the assistant can call them during a conversation.
Tools picker
Every tool the chat can use — built-in TIA tools, tools from MCP servers, and tools from other extensions — appears in the chat tool picker, grouped by source. Each tool has an enable toggle so you can narrow what the assistant is allowed to use for a given conversation.
Tool approvals
Tool calls that change your project are gated by a per-tool approval level you set in the Tool Approvals section of the Agent Customizations editor. Each managed tool (TIA, EPLAN, Visual Context and Commissioning Report) has the same choice of Ask Every Time, Allow in This Session, Always Allow, or Always Deny, with no exceptions: at Ask Every Time the in-chat confirmation appears before each call, and a wider level lets the tool run without asking. Your choice takes effect immediately for that tool's in-chat confirmation prompt. Always Allow and Always Deny persist across window reloads; Allow in This Session resets to Ask Every Time after a reload; a Reset All link clears every override.
Terminal command risk notes
When the assistant proposes a terminal command that needs your approval, Studio can add a one-line, AI-generated risk note to the confirmation — for example "deletes files recursively" or "downloads and runs code". Turn it on in Settings: search for AI Risk Analysis (under the terminal tool's chat settings); it is off by default. The note is advisory only: it never changes whether a command needs approval and never makes a command run on its own — the command you see and the usual approval still apply. To produce the note, the command is sent to your selected AI model, so it appears only while you are signed in to an AI provider; if the model is slow or unavailable, the confirmation simply shows no note.
License gating
The Siemens reference guides that the assistant draws on are license-gated: they feed the model only while your account holds an active AI-chat entitlement (Basic, Pro, Pro+, Enterprise, or an active Trial). With an entitled license they simply work — no key, dialog, or extra step. On sign-out or a tier downgrade they stop resolving. AI Chat itself is available from the Basic plan; the same entitlement governs the bundled reference material.
Diagnostics
When a tool call or an MCP server misbehaves, use the Output panel at the bottom of the window: pick the AI chat or MCP channel from the channel dropdown to see the relevant log stream. The Problems panel surfaces issues raised while a project is connected.