Reference

Full capability list

Everything Coilmind can do, grouped by what you’re working on. If it’s on this list it works today, and where something has a limit, the limit is listed next to it.

Project analysis

  • —Read and explain FBs, FCs, OBs and DBs across the open project
  • —Cross-references: who writes to a tag, who calls a block
  • —Call graph and structural overview
  • —Dashboards alongside the chat: architecture, project health, instance usage, reusable blocks, project statistics
  • —Flag logic inconsistencies
  • —Document blocks or the whole program from the LAD and SCL itself: purpose, interface and behaviour of each block
  • —Draw the logic as Mermaid flowcharts
  • —Export the documentation as a Markdown (.md) file
  • —Read screenshots pasted into chat
  • —Read your own documents attached to the project, and search across them
  • —Keep a project specification the assistant reads before it writes

AI models

  • —Cloud model by default: AI included in every plan, no key to manage and no tokens to buy
  • —Your own key: point Coilmind at any OpenAI-compatible endpoint
  • —Local model through Ollama, running on your own machine; nothing leaves the PC
  • —All three options on every plan
  • —Cloud inference runs at DeepInfra and Fireworks in the US; no copy of prompts or project is kept on our side
  • —Local models are weaker than the cloud ones, and how capable they are depends on your GPU
  • —The templater and the batch generator use no model at all

PLC programming

  • —Generate and import LAD
  • —Generate and import SCL
  • —Read STL
  • —Rewrite legacy STL into SCL or LAD
  • —Clone blocks with recursive tag substitution
  • —Self-correcting compile loop
  • —Validate a block without writing it: topology, schema and semantics as a dry run
  • —Preview a network before it is applied

Batch generation (no AI)

  • —Queue one template across a list of blocks and create them in a single pass
  • —Allocate tags automatically and walk start addresses
  • —Report of everything created, and a history of previous runs
  • —Fully deterministic, with no model involved at any point

Tags, types and project structure

  • —PLC tags: add, update, rename, delete
  • —Tag tables and tag groups: create, organise, search across the project
  • —UDTs: create, modify and delete types and type groups
  • —Rename a block and relink every caller
  • —Move blocks between groups; list and manage block groups
  • —Block metadata (title, author, comment) in every project language
  • —No snapshot and no compile gate on tag, tag-table or UDT operations: undo those in TIA

Legacy and migration

  • —Read STL networks and rewrite them into SCL or LAD
  • —A migration cookbook the assistant follows rather than improvising
  • —Anything that cannot be translated cleanly is surfaced, never silently guessed
  • —Prove a rewrite by running the original’s tests against it (needs PLCSIM Advanced)

Testing

  • —Declarative .test.json suites: inputs, cycles, assertions
  • —Execute against PLCSIM Advanced V3–V6
  • —Scaffold a suite from a spec
  • —Live progress while a run executes, and cancel mid-run
  • —Download to the simulator only. Fail-closed: it will not reach real hardware
  • —PLCSIM Advanced is a separate Siemens product; the rest of Coilmind does not need it

HMI and WinCC Unified

  • —Create screens and screen groups
  • —Create, modify and delete screen items
  • —Change any property of a screen item: position, size, colours, text, the tag it shows
  • —Set dynamizations on items
  • —Write JavaScript for item events and script dynamizations
  • —List and describe available item types and library types
  • —HMI tags, tag tables and groups: full create, update, delete
  • —Compile the HMI
  • —Monaco editor with IntelliSense from the Siemens V21 TypeScript definitions
  • —Project-wide search across scripts and tags; SyntaxCheck through the Siemens API
  • —Does not create HMI devices; faceplate types cannot be created through Openness by any tool

Working across the project

  • —Switch between PLCs and HMI devices inside one project, in one conversation
  • —Assistant memory: project conventions and preferences survive the session
  • —Changelog of every change the assistant made, kept with the project
  • —Take whole tasks from your own agent (Codex, Claude Code and other clients that connect to a local MCP server), with the same validation, compile gate and snapshots
  • —The conversation between your agent and Coilmind is visible live in a separate chat

Translation

  • —Titles, comments and tag descriptions, in the context of the logic
  • —EN · DE · IT · ES · FR · RU · ZH

Templates (no AI)

  • —Mark replaceable tags and parameters, type new names
  • —Fully deterministic, works offline, no model required

Setup, safety and rollback

  • —Openness access configured automatically on first run: one Windows prompt, no manual whitelisting
  • —Attaches to a project already open in TIA Portal; it does not open projects for you
  • —Local snapshot before every block and HMI screen change
  • —One-click restore from history
  • —Compile gate after every block import: if it doesn’t build, the change isn’t applied
  • —Up to 14 days of offline work before the licence needs a short refresh
Start your 14-day free trialMissing something you need? Email support@svghmi.pro and ask. Most of the roadmap comes from emails like that.