The Desktop Agent

SphereHarness lives beside your CAD system rather than inside it. A small floating balloon sits on top of whatever you are working in, one keystroke brings up a place to type, and the same conversation reaches every CAD system you have running — including two at once.

This replaced the chat panel inside the CAD window. That panel still exists and is switched off by default; see the legacy panel at the foot of this page. If you are reading older notes that tell you to open a task-pane tab, this page is the current answer.

Starting it

The agent runs in the Windows system tray. It starts:

  • When you click Open Desktop Agent on your CAD system's SphereHarness ribbon. (In AutoCAD the button is Desktop Agent, and the commands COPILOTAGENT and COPILOT both do the same thing.) If the agent is not running yet, the ribbon starts it.
  • When you launch it yourself, or with Windows if you tick Start with Windows in Settings → Agent. That is off by default, as is Start hidden (tray only).

Once it is running, closing any of its windows only hides them. Exit on the tray menu is the only thing that stops the agent.

The SphereHarness agent icon and status in the Windows system tray
The agent in the tray. Hover it to see whether it is connecting, connected, stopped, or not signed in.

The balloon

A small round disc that floats above every window. It is blue when you are signed in and grey when you are not, and the glyph on it is a chat bubble — or a microphone when voice is switched on, or a padlock when you are signed out.

  • Click it to open a place to type.
  • Drag it anywhere. On release it snaps to the nearest screen edge, and remembers where you put it — per monitor arrangement, so docking and undocking a laptop does not lose it.
  • Right-click it for the menu: Open chat (Sign in… when you are signed out), Macros…, Markup board…, Settings…, Channels and services…, Hide balloon (tray keeps running), Sign out and Exit.

Three things are shown on the disc itself, so you can read the state without opening anything:

  • A red unread badge when messages have arrived from one of your channels.
  • A health dot at the bottom-left — hover it for CAD sessions healthy, No CAD session connected, or A CAD session crashed.
  • A macro glyph at the bottom-right that opens the macro library, and turns red while a recording is running. See Macros.

Hidden the balloon by mistake? The show/hide hotkey brings it back, and Settings → Agent has Reset balloon position for when it ends up on a monitor you no longer have.

The quick ask bar

One line, floating near the top of the screen, for the question you want to ask without leaving what you are doing. It is what a balloon click opens by default.

  • Ctrl+Alt+Space opens and closes it from anywhere. You can change that in Settings → Agent (Show / hide hotkey), or clear it if it clashes with something.
  • Enter sends, Shift+Enter is a newline, Esc closes it, and Ctrl+Enter opens the full chat window instead.
  • The answer appears inline underneath, with a button to carry on in the full window.
  • It has buttons for the markup board and, when voice is enabled, for talking — the talk hotkey is Ctrl+Alt+V by default.
  • What you say is not sent straight away. The words go into the box, after anything you had already typed, with the cursor at the end — correct them if the recogniser misheard, then press Enter or Send. If you want hands-free sending, tick Send at once in Settings → Voice (off by default).
The quick ask bar and the chat card are two views of one conversation. Ask in the bar, open the window, and the exchange is already there.

The chat card

The full window: the conversation, the input box, and the controls. It is what the CAD ribbon button opens, and what Ctrl+Enter from the quick ask bar opens. Drag its edge to resize it, or use the card menu's Dock submenu to pin it to the left, right, top or bottom edge of the screen instead of leaving it floating.

Moving and resizing the chat card

  • Resize it from any edge or corner, at any time. It cannot be made smaller than 260 × 220 or bigger than the screen's work area (the screen without the taskbar).
  • Move it by dragging the blue title. Dragging it against the top of the screen makes it fill the screen; against the left or right edge, Windows snaps it to that half. It stays resizable in every one of these states.
  • Get its own size back by double-clicking the title, or by dragging the title away from the edge. Double-clicking the title of a card at its own size makes it fill the screen.
  • It remembers its size and where it was on each monitor, and brings a remembered position back onto the screen if that monitor is gone.
  • Ask the chat: “make this window bigger”, “resize this window to 600x800”, “move it to the right”, “fill the screen”, “put it back”, “dock it on the left”, “don't keep it on top”. The AI knows the chat card is its own window and acts only on that window. Keep the chat window on top of other windows is also a setting (Settings → Agent).
Ask “what can you do in this app?” and the AI describes SphereHarness itself: its windows, the AI and model pickers, the working folder, the licence, channels, macros and support.

Typing

  • Enter sends; Shift+Enter starts a new line.
  • Type / to open the command palette — /help, /clear, /compact, /stop, /pause, /usage, /verbose, /context, /think, and in the card also /restart, /settings, /sessions, /model and /setupwizard.
  • The paperclip attaches a file to the message.
  • Replies are rendered as formatted text — headings, lists, tables, and code blocks with a Copy button. Screenshots SphereHarness takes appear inline.
  • The chat also draws charts, flowcharts and sequence diagrams, option buttons you click to answer, and collapsed sections for long logs — on your PC, with no internet access. See What the chat can show.
The Send button becomes Stop only when the box is empty. While a turn is running, the blue Send button turns into a red Stop — but only if you are not typing. If you have started a follow-up message, clear the box to reveal Stop.

Choosing which AI answers

Just above the input box there is a small AI drop-down — which AI answers in this chat: your own API key, or a coding agent — with a gear beside it that sets the chosen one up. The quick ask bar has the same choice on its menu.

  • Your own AI key — a key from OpenAI, Anthropic, DeepSeek, Google, xAI, Kimi or any OpenAI-compatible provider, or a local Ollama. The key is stored encrypted for your Windows account on this PC only (%AppData%\CADCoPilot\api-keys.dat); it is never sent to SphereHarness.
  • Coding agent on this PC — a coding agent installed and signed in on your PC, such as Claude Code, Codex or Gemini CLI, under your own agreement with its vendor. The first time you pick an agent, the chat shows a one-time notice: it runs headless, some vendors require an API key when another program drives their agent, and the feature depends on the vendor's continued support.
  • SphereHarness server — listed only when your account is offered it. Your requests pass through our server to the AI provider we use for that service and are not stored; it runs on a token allowance (a free allowance, then tokens you buy). The line shows your tokens left and warns when they run low; at zero the choice stops and your own key or coding agent keep working. See how tokens work and what we store.

Either way the conversation goes straight from this PC to that AI; SphereHarness does not run AI on its servers for these two choices. On a new install the chat opens with a card that offers them. If the optional SphereHarness server choice is offered to your account you can pick it instead (see the list below); an install that used the former server option and is not offered it shows the card and answers nothing until you pick one. Where your data goes: What goes where.

Each option tells you whether it is ready to answer, and that verdict is the backend's own pre-flight check — the same check it makes when you press Send — rather than the picker's separate opinion of what “configured” means.

Setting up a coding agent is not always one click. Gemini CLI can be installed for you; Claude Code and Codex need their own command-line tool installed and an interactive login first. The set-up window does what it can and then tells you the command still to run, rather than reporting success and failing at send time.

The setup wizard

The setup wizard walks you through Your AI, then Teams, WhatsApp and Telegram (optional), then a test. It opens by itself once, after the first sign-in on a new install when no AI can answer yet; after Finish or Not now it never opens by itself again. Open it any time from the tray (Setup wizard…), the status window, the small Setup wizard button just above the message box, or by typing /setupwizard. It resumes at the first step you have not finished.

  • Your AI lists your own API key and every coding agent this PC knows, each Ready, Installed - check the sign-in, Needs sign-in, Needs Node.js or Not installed. For a missing agent it shows the vendor's own install command and Open PowerShell and install: a visible PowerShell window prints the command and runs it. SphereHarness never installs anything hidden and never asks for administrator rights itself; Node.js installs for all users, so Windows asks you to approve it (UAC); on a PC without winget (Windows Server 2016 / 2019, some LTSC) — or when winget fails — the wizard downloads the official LTS .msi from nodejs.org, checks its SHA-256 against nodejs.org's SHASUMS256.txt (a mismatch is deleted, never run) and starts it. The wizard watches the install, refreshes PATH from Windows, and turns the row to Installed by itself.
  • Before an agent is started for the first time — even for the check — the wizard shows the same one-time notice as the chat. Check sign-in opens a session without sending a message; Open the sign-in window runs the agent's own sign-in command (claude, codex login, …) where you sign in. SphereHarness never sees your sign-in.
  • When an agent will not start, the wizard names the first missing link — Node.js, npx, the vendor's program, the npm registry — with the command that fixes it. “The Claude adapter's download is damaged - Repair” removes only the incomplete npm-cache\_npx\<number> folder an interrupted first download left behind (never the npm cache); the chat repairs it once by itself too.
  • Teams, WhatsApp and Telegram show Connected, Set up, Organisation, Paused or Not set up; Set up… opens the channel dialog. Teams goes through the SphereHarness Teams app (relayed, never stored; an administrator may need to upload the app once); WhatsApp uses your own Twilio number and Telegram your own bot. See Channels.
  • Done sums it up and Test it sends one short message in the chat.
An AI that is not ready never takes a message. While nothing can answer, an amber line above the message box offers Start the setup wizard, Retry and — when another AI is ready — Use it; the red Not ready status line has Set up…. A message sent anyway goes back into the box with the reason. If the AI in force is not installed and you did not choose it yourself, the chat switches to a ready one and says so. The coding agent's own error output is kept in %LOCALAPPDATA%\CADCoPilot\logs\coding-agent.log.

Choosing which model it uses

Beside the AI drop-down is a Model drop-down. It lists what the AI you picked actually offers — a coding agent's list comes from the agent itself, and your own API provider's comes from the list its /models endpoint returned when you pressed Fetch models in Settings. Nothing in that list is a name we keep here, so it cannot go stale.

Leave it on (the agent's own default) to let the agent keep whatever model it is configured for. Pick one and it applies to the agent already running: your next message uses it, with no restart and without losing the conversation you are in.

The choice is remembered per agent, so a model you chose for one coding agent is never asked of another, which that other agent would simply refuse. You can also type /model in the chat to see the model in force and what the agent offers, or /model <id> to change it.

The box is editable, so you are not limited to the list. A list can lag — a model released after the adapter was built, an id your account is allow-listed for, a deployment name on your own endpoint. Type it and press Enter, or click away, and it is sent exactly as you typed it. If the agent takes it, it is used from your next message and appears in that agent's list next time marked (added by you), so you can tell your own entries from the ones the agent named. If the agent refuses it, the chat says so in the agent's own words, you stay on the model you were already using, and the refused id is neither kept nor added to the list.

The Model boxes in Settings → Coding agents & MCP work the same way and share the same store, so the chat card, Settings and /model never disagree about which model is in force.

Choosing the coding agent's working folder

Beside the Model drop-down is a small folder button showing the name of the folder your coding agent works in — its working directory, where its own file and shell tools start. Hover over it for the full path. Click it for a short menu: your recent folders, Default, and Choose folder…, which opens the standard Windows folder picker. On a narrow card the button shows only the folder icon, so the AI and Model boxes keep their room.

  • Default is exactly what the agent used before this button existed: the folder set in Settings → Coding agents & MCP → Working folder for launched agents, or Documents\CAD CoPilot Projects when that is empty. Nothing changes until you choose a folder.
  • The choice is remembered per coding agent, like the model, and survives a restart.
  • A coding agent fixes its working folder when a conversation starts, so changing it starts a new conversation with the agent. The chat tells you in one line — Working folder changed - the next message starts a new Claude Agent session in … — and your next message goes to the agent in the new folder. The agent itself keeps running.
  • If a folder you chose has since been deleted or renamed, the chat says so and the agent uses the default instead; it is never recreated.
  • Only a folder on this PC can be chosen. A network (\\server\share\…) folder is accepted only while it can be reached, and a whole drive such as C:\ is refused: the agent's file access stays within one folder.
  • Choosing a folder grants nothing new. The coding agent's own permission settings apply there exactly as they do anywhere else, and when it asks permission for something that names a path, the card also shows the working folder.
Lean coding agent (Settings → Coding agents & MCP, on by default): Claude Code started by SphereHarness ignores your global Claude Code setup — your user settings, plugins, your own MCP servers and your global CLAUDE.md — so each conversation starts with far fewer tokens (measured: about 40,000 down to 23,000 on the first message). The working folder's own CLAUDE.md and project settings still apply, Claude Code's Chrome tools stay on (with Lean off too), and your Claude Code sign-in is untouched. Add MCP servers: put them in your working folder's .mcp.json (or .claude\settings.local.json there) in Claude Code's own format; they are passed to Claude Code from the next conversation, and a file or server that cannot be used is named in the chat and skipped. Untick Lean if you want your own plugins and user-level MCP servers in the chat; it applies from the next conversation.
Your own AI key has no working folder. Its tools work on the CAD documents you have open, not on files in a folder, so the folder button is greyed out while your own key is chosen, and its tooltip says why. In Non-CAD mode the commands it runs start in the Default folder.
The SphereHarness server choice appears only when your account is offered it. When it is not, SphereHarness does not run AI on its servers, and if a conversation still names it, the chat asks you to pick your own key or a coding agent before it answers.

Non-CAD mode

The Non-CAD mode check box, beside Target at the top of the chat card (also in Settings → Agent), makes SphereHarness behave as if only the desktop agent were installed: the AI gets no CAD tools and no CAD instructions, so each message is roughly half the size — fewer tokens, lower cost, faster answers. It applies from your next message, with no restart; a coding agent starts a new conversation, as it does when you change its working folder. While it is on, the Target picker and the macro library button are disabled.

  • Still available: any window on the PC (desktop_ui: windows, UI elements, keyboard, mouse, screenshots, starting and ending programs), SphereHarness's own browser, knowledge, memory, skills, company connections, the markup board, the agent's settings and support.
  • Windows of programs running as administrator (installers, admin tools): the AI asks you first, then Windows shows its consent prompt; only SphereHarness's Windows UI helper is elevated.
  • Commands and scripts with your own AI key: the AI can run PowerShell, cmd, Python (one already installed on the PC) or Git Bash, as you and never as administrator, in the working folder. Every run shows the code on an approval card — Allow once, Allow … this session or Reject — stops after its timeout and returns the output and the exit code. A coding agent uses its own shell.
With no SphereHarness CAD add-in installed on the PC, SphereHarness is always in non-CAD mode: the box is shown ticked and greyed out, and its tooltip says why.

The header

The header shows who you are signed in as and what is answering, so you always know whether you are on your own key or your own coding agent.

The icon buttons open the markup board, the macro and variation library, the notification bell, and a Settings and more menu holding Settings…, Channels and services…, Refresh CAD sessions, the Dock submenu, the coding-agent options, Send feedback… and Sign out. The Close button hides the window; it never exits the agent.

The notification bell

A bell in the header, with a count on it when something has arrived that you have not read. It is how a reply to one of your support requests reaches you without your going to look, and it is where announcements from your partner or from the SphereHarness team appear.

  • Click it for the last 20: the title, the text and when each arrived, with Mark all read at the bottom. Clicking one marks it read, and then a support notification asks the chat to show that request (so you read its current state, not the notification's frozen text) while anything carrying a link opens the link.
  • A tray balloon is raised as well, so a notification is not missed because the card is behind your CAD window. One balloon per check, never one per notification - five answers arriving together are one interruption.
  • It is checked every 60 seconds, and only while you are signed in. The first check after you sign in sets the count but raises no balloon, because a balloon means "this arrived while you were sitting here".

Every message you send gets a progress block underneath it, in the conversation itself. Its top line always shows and refreshes every second while the turn runs — how long it has been going, how many CAD calls it has made, how many of those are dialog-driving (driving the CAD's windows rather than its API), and the last operation with how long ago it ran. Clicking that line opens or closes the list below it.

The list holds one line per call, oldest first, naming the operation, what it was given and which CAD session it ran on, then how it ended and how long it took — CreateExtrusion distance=20 isCut=false · Inventor #9184 — completed 2.5 s. A ⌨ marks a dialog-driving call and a ✋ a permission ask; sub-agents get their own line too. The list has a fixed height and scrolls, so a turn of two hundred calls never pushes the conversation off the screen: it follows the newest line while the turn runs, and stays put if you scroll up in it.

The block is open while the turn runs and collapses to its one-line summary when it ends — 96 tool calls · 78 CAD · 34 dialog-driving · 15 min 0 s — which a click re-opens. A turn that called no tool gets no block at all. A turn that is mostly dialog-driving is SphereHarness working around a capability it does not have as an operation; it says so in one sentence when it starts. The model's reasoning can be shown too — Settings → Agent, off by default.

Messages that reach SphereHarness from a plain-text channel (Telegram and the like) have no block to show, so those answers still carry the [SetDialogElement ×8 — completed] tool lines inline.

The Target box

Under the header is a Target drop-down. It decides which running CAD session this conversation works in.

  • (let the model choose) is the default. Each request is routed by what it names — which is fine when only one CAD is running, and fine when you say “in NX”.
  • A named session — for example Siemens NX #208696 — pins the conversation to that one running instance. The document is whatever that instance has open.
  • Launch … entries appear greyed for CAD systems you have installed but are not running. Choosing one is an action, not a selection: it starts that CAD, and the box goes back to (let the model choose).
Pick the Target when you have more than one CAD open. If a request names no session and more than one could serve it, SphereHarness refuses and lists them rather than guessing — which is correct, but costs you a round trip. Picking first avoids it. It also saves tokens; see Spending fewer tokens.

The tray menu

Right-click the tray icon for: Open chat, Open the full chat window, Show balloon, Ask my coding agent…, Setup wizard…, Settings…, Open status…, Start relay / Stop relay, Sign out and Exit. Double-clicking the icon opens the chat.

Hovering the icon tells you the connection state — starting, connecting, connected, stopped, or not signed in.

Permission cards

Before SphereHarness does anything destructive to your model — and before it changes its own sensitive settings — a card headed Permission needed appears on whichever surface is open.

  • Allow once — this one, now.
  • Always allow … this session — that operation, without asking again, until the agent restarts.
  • Reject.

Each card belongs to one CAD session and says so, with a count when more than one is waiting (“Siemens NX #208696 · 1 of 2 waiting”). The cards sit in a strip above the conversation (never down in it), each with its own countdown (“4:32 left”): you have five minutes from the moment a card becomes visible. While one waits, the chat window is restored, kept on top and flashes in the taskbar, a sound plays (Settings → Agent → Play a sound), and the tray notification shows the same Allow once / Reject buttons — answer there, or click the notification to bring the chat window forward. A card nobody answers is refused (“nobody answered within five minutes”) and nothing runs. Every card appears there — the setting changes and channel saves the AI asks for included. When the quick ask bar is open it shows the same card (answer in either place); closing the bar refuses nothing. A request that was withdrawn leaves a one-line note with a Dismiss button rather than a stale card.

When nobody is at the PC. A request that came from a channel, the local chat API or another program on the PC does not wait for the card: the AI gets needs approval at once, nothing runs, and its answer says what is waiting. The card stays on the desktop; once you allow it there, the same request runs without asking again. Deleting or closing something the conversation did not create always needs you, even in unattended mode.

Long jobs continue by themselves (1.3.7). You never have to type continue: a request runs until it is done. The time budget (Settings → Agent → Time budget per chat turn, default 15 minutes) is now a checkpoint: at each one the chat — and Teams, Telegram or WhatsApp when the request came from there — gets a line such as Checkpoint: slice 2 done · 30 min so far · continuing automatically · running: sub-agent 'nest parts', and the work goes on. Nothing is interrupted at a checkpoint: the coding agent's sub-agents and background tasks, and a screen recording the request started, keep running. The balloon's busy ring tooltip shows the same: time, slice, “continuing automatically” and what is running. When the AI's own turn ends before the work does, the request continues in a new turn of the same conversation by itself.

A long job stops by itself only at a safety limit, and then it saves the open document(s) in the job folder and says exactly why and how to resume (reply continue): its total time (Longest a job may run, default 240 minutes), the same tool error coming back (Stop a job when the same tool error comes back, default 5 times) or nothing happening — no reply, no tool or CAD call running, no recording — for Stop a job that stalls (default 20 minutes). These settings, and the switch Long jobs continue by themselves (on; off brings back the old behaviour where a request ends at its budget and you reply continue), can only be changed by you, not by the AI.

Stop and pause. Stop, or /stop, ends everything the request started: the running turn, its sub-agents (named with their last step), its background tasks (the agent is told to end them) and a screen recording it started (the file is finalised and named). /pause — or just pause as the whole message while it runs — does the same, saves the open document(s) and says what was left; reply continue to resume. If a sub-agent ever ends before it finished, the chat says so at once with its name and last step, and the AI is told it is not running.

When your chat is being answered by your own coding agent rather than by SphereHarness's model, that agent supplies its own permission wording, so the buttons will read however it phrases them.

Settings → Permissions

If you do not want unattended mode but are tired of one kind of card, open Settings → Permissions. It lists every permission the agent can ask for, one row each: the box Allowed without asking (ticked = no card for that kind; unticked = the card asks you with Allow once, Allow for this session or Allow always, which ticks the box), where the answer comes from now (Always, this session, default or unattended mode) and Revoke. A change takes effect at once.

  • The rows: deleting, closing or undoing in a CAD document; running commands (PowerShell, cmd, Python, bash — also how software is installed); ending a program; removing knowledge files; deleting macros; removing organisation skills; writing through company connections; controlling SphereHarness's own windows and settings; controlling other programs' windows (allowed by default); the coding agent's own requests; the coding agent reading and writing files outside its working folder; adding or changing company connections; running the Windows UI service as administrator; filling a saved sign-in (set per sign-in on the Saved sign-ins page).
  • The rows that always ask you are listed too: sensitive settings changed from chat, commands that touch SphereHarness's own settings, licence / sign-in / security windows, documents the conversation did not create, foreign solvers, add-in registration fixes, variation sets that rewrite reused files, the Teams identity check — and SphereHarness's own approval cards, which only you ever answer.
  • Ticking a risky row (⚠: deletes, commands, ending programs, company systems, the coding agent, administrator rights) shows a short warning you confirm first.
  • While unattended mode is on, every row you can tick is allowed without asking and shown ticked and greyed — company connection changes, administrator mode and the coding agent's own requests included.

Unattended mode has one meaning. Unattended mode: SphereHarness answers every permission without asking (except its own approval cards). Switch off any time. That includes a coding agent's own requests. Switching it on at the PC (the shield button, /unattended or the Settings checkbox) shows that sentence with one choice: Turn on unattended mode or Keep asking me (with a coding agent it reminds you of your own agreement with its vendor). The older “CAD approvals only” choice is gone; a PC that had unattended mode on now has the full meaning. /unattended session, /unattended always and /unattended off sent from your linked, verified Teams, Telegram or WhatsApp account or through the chat API switch it on this PC without a dialog: the reply says what was switched and how to switch it off, and the switch is written to the unattended log and shown as a tray notice on the PC.

The AI can drive SphereHarness's own windows. Ask “open Settings and turn on the dark theme” or “untick Allowed without asking for commands”: the AI may click and type in its own Settings, chat card, status window, setup wizard, board and browser window — their Sign in / Accept buttons too — and says first what it will change. Unless the row Control SphereHarness's own windows and settings is ticked or unattended mode is on, a card asks you once. Licence, sign-in and consent buttons of other programs, and of web pages shown in SphereHarness's browser window, still need you.

Only two limits in SphereHarness's own windows, in every mode:

  1. Approval cards and permission confirmations — the cards' Allow / Not me / Reject buttons, the unattended-mode consent, the administrator-mode dialog, the channel risk dialog and a risky permission's warning. The AI can never press them, so a risky tick, unattended mode or administrator mode always ends at a confirmation only you can answer. (A click you make on one in the second while the AI is driving the desktop is ignored — click again.)
  2. Filling a saved password into SphereHarness's own windows — refused, because a secret in one of its boxes could be read straight back.

Other windows on your PC

Outside the CAD, the chat can work with any window on your PC: a browser, Explorer, an installer, a licence tool. It can list and focus windows, read their text and controls, click buttons, type into fields, fill a dialog by its labels, press keys, click, drag and scroll the mouse, move, resize, maximise and restore windows, take screenshots, read the screen layout, put text on the clipboard and read it back, start programs and list running ones. Ask for these in ordinary words, for example “drag Notepad to the left half”, “maximise the browser” or “copy this text to the clipboard”.

  • Ending a program always shows a Permission needed card that names the exact program and its process id. Nothing ends until you allow it. The chat can never end SphereHarness itself or the helpers it runs on.
  • Mouse and clipboard actions run without a card, and each one appears in the turn's progress. For the clipboard, the progress shows only how many characters were copied, never the text itself.
  • These actions need a valid SphereHarness licence on this PC. Without one, the chat says so and does nothing.
  • They run with your own Windows rights. Working with programs that run as administrator is off unless an administrator allows it; see Administrator-level operation.
  • The helper that does this work, the Windows UI service, answers only SphereHarness. Other programs on the PC cannot use it.

Opening a program by its Start-menu name

Ask “open Vault Professional 2026” or “start Calculator”: the chat starts the program by its Start-menu name (desktop_ui action launch_start_menu_app). It does not type into Windows search — Windows 11 draws the search results in a way the Windows UI service cannot read.

  • The name is matched against the Start menu's own entries only: its shortcuts (for all users and for you) and the installed Store apps. The program is started from its shortcut by the Windows desktop, with your normal rights, as if you had clicked it. No file path or command line can be started this way.
  • One clear match is started. Several matches (“Vault 2026” = Vault Basic 2026, Vault Professional 2026, the Vault tools) start nothing: the chat gets the list and asks you, or calls again with the exact name (with its Start-menu folder when two entries share a name). No match names the nearest entries.
  • CAD programs (Inventor, NX, Solid Edge, AutoCAD) are started by the chat's CAD launcher (cad_launch), which connects the SphereHarness add-in.
  • Command Prompt, PowerShell, Windows Terminal, Registry Editor, Task Manager and the management console are refused, as for every other way of starting a program.

Several CAD systems at once

The agent is one process talking to every CAD add-in that is running, so you can have Inventor and NX open together and address each in turn. Two things follow:

  • Work on one CAD does not interleave with work on another — each CAD has its own queue, and a macro run holds its CAD for the whole run.
  • Ambiguity is refused rather than guessed. If it is not obvious which session you mean, you are asked, or you set the Target.

Asking for help, and telling us what you think

Two different things, both from the chat card.

  • A support request opens a thread somebody answers. Ask for one in the chat, and a diagnostics zip - the SphereHarness logs, your CAD's own recent logs and crash dumps, what is installed and which add-ins are loaded - can be built and attached as the request is created. Collecting uploads nothing on its own. You can attach any other file you already have, ask what the status of your requests is, add a message or another file to one that is open, and close it when you are done. The reply comes back through the bell. Full details on Support requests.
  • Feedback is what you think of the product, not a question. Send feedback… on the Settings and more menu opens a short dialog: your words, an optional Rating from 1 to 5, and Send. It gets no request number and nobody is obliged to reply, which is why it is the right place for "this is awkward" rather than "this is broken".
Collect my Inventor logs and open a support request about the crash, attach them
What is the status of my support requests?

“Update available”

The agent checks once a day, and when you sign in, whether a newer build has been published. When there is one you get a tray balloon headed Update available:

“SphereHarness 1.3.0.0 is available (you have 1.2.0.0) — download it from the portal.”

The same line is on Settings → Account, with Open the downloads page next to it; when you are current it reads Version 1.2.0.0 installed. instead. Your CAD add-ins mention it once per session too, on the status bar or in their own log.

  • One balloon per new version, not one per check — it does not nag.
  • Nothing downloads or installs itself. The only action offered is opening the portal page.
  • The check runs only while you are signed in, and five add-ins plus the agent starting together make one check between them.

Where the installers are and which ones you are offered: Downloads.

What the installer put on this PC

You did not install the agent separately — every CAD system's installer carries it, along with the shared configuration and knowledge folder. That is why installing for a second CAD system does not give you a second agent: the existing one is reused, and upgraded if the newer installer carries a newer copy. One agent, all your CAD systems, one update to keep track of. See Installing SphereHarness.

The legacy in-CAD chat panel

The chat panel docked inside the CAD window is still in the product but switched off by default. It can be turned back on in the SphereHarness Settings application, on the SphereHarness page: Show chat panel inside the CAD (legacy).

  • It is one setting for every CAD system, not one per CAD.
  • It is read when the add-in loads, so restart the CAD application after changing it.
  • Turning it on changes where you can type. It does not change how messages are routed — that still goes through the desktop agent either way.

With it off, a CAD system that has the add-in installed shows a short panel saying that chat has moved and that the session is still connected, plus the ribbon commands.

See also