Grounding for AI workers (BrainstormAI, DesignAI, WordSmithAI, ImageGenAI) when a user asks about the platform itself, and — verbatim — the public page at /support/ai-reference. Verified against the actual codebase, not assumed. Contains no internal implementation detail, no keys, no user data, and no roadmap-only information.

This file is PUBLISHED. It states facts only. Anything that instructs the model rather than describing the product belongs in platform-reference-directives.md, which is loaded alongside this file for every request but never served to a browser. Instructions used to live here and were published with the facts — including one that named the two roadmap ideas it was forbidding any mention of.

VisionAI Workspace is a web-based AI project-management platform — not a tool for building other apps or websites. It turns an idea into a plan an AI team can actually carry out — the user stays in control of every step.

The real, currently shipped capabilities of the platform. If you're not certain something on this list covers, ask rather than assuming.

• Orbits — a goal-driven workspace; everything else lives inside one
• Orbit Modes & Phasing — Launch/Pulse/Cycle orbit types; Launch orbits additionally get a phase roadmap, see below
• Tasks — units of work inside an Orbit, each assigned to one AI worker
• The AI team — BrainstormAI, DesignAI, WordSmithAI, plus ImageGenAI (image generation), see below
• Energy states — the task/artifact lifecycle, see below
• Artifacts & Versions — the deliverable a worker produces, with a version history and accept/revise flow
• Follow-up suggestions — AI-proposed next tasks offered after accepting an artifact, aligned with the orbit's current phase for Launch orbits
• AutoLoops — scheduled recurring automation on an Orbit: Momentum Loop (produces a fresh deliverable each cycle) and Insight Loop (read-only recurring analysis), see below
• Focus Week — a weekly scheduling view for booking time against specific tasks, with Google Calendar export; crew can be invited to a session and accept or decline, and an invited session can carry a Google Meet link, see below
• Shared workspaces — a workspace owner can invite other people into their workspace by email, choosing which orbits each person can reach, see below
• Crew messages — in-app messaging between people in the same workspace: a room per orbit, plus direct messages, see below
• Orbit Sync — a live session on an orbit: one person leads and everyone following sees the view and task they open, with optional voice chat; see below
• Orbits Map, places and work orders — saved customers/places on a map; work orders (field visit or remote) created from a task, assigned to a crew member, and worked through set steps; My work orders; place history, see below
• Inventory — parts and equipment with counts at each stock location (a warehouse or shop on the map, vans, or a virtual location such as Main stock or Made to order); add items one by one or import a spreadsheet; set counts and record stock received; low-stock flag at a reorder point; add an item from a photo (AI fills in what it can see); cost and sale price visible only to the owner and Orbit Managers. Each van is assigned to one crew member, who sees only that van ("My inventory"), uses its parts on their work orders, and asks for a count fix that a manager approves; Orbit Managers load vans from a warehouse. Each item keeps a photo and a QR code (printable labels); "Scan a part" identifies an item from a photo. Low-stock alerts (an alert level set per item) reach the owner and Orbit Managers. Restocks are work orders on each orbit's one "Restock from warehouse" task: planned by a manager, requested by a driver and approved, or made automatically the evening before a van's scheduled visit from its "should carry" levels; the driver records what they load and it moves warehouse-to-van on the spot. Each work order shows a parts and billing summary (cost, sale price, margin) to the owner and managers.
• Work order requests — crew who are not Orbit Managers request a work order through the same wizard; the orbit's managers or the owner approve or decline it
• Search — one search beside the notification bell for customers, or for everything (orbits, tasks, artifacts, work orders, crew, sessions)
• The Orbit Canvas — the visual, node-based map of an Orbit's tasks
• List View — a flat, cross-orbit list of tasks
• The Vault — an archive of saved/accepted AI outputs
• Momentum Packs — one-time paid boosts (Orbit Boost, Momentum Refresh, ImageGenAI Burst, Task Boost) for a Starter-plan limit that's genuinely in the way right now; not required, every pack has a real free alternative — full pricing below
• Badges — an in-app achievement system tied to platform usage
• Reference files — attaching a file to a task or AutoLoop as grounding context for that AI worker
• AI tone (Casual / Professional / Creative) — two independent controls, not one account-wide setting; see the AI Tone section below
• Multi-language support

This list is the complete set of real, currently shipped features. There is no API or webhooks, no custom task states, and no integrations beyond those named above. Pro and Enterprise are described in their actual, limited confirmed scope in the Plans section below.

This paragraph used to read "there are no collaboration/team features", which stopped being true when shared workspaces and crew messages shipped. Both are real and both are described below — do not tell a user this product is single-player.

Same two-layer format as Energy states and AutoLoop above: technical definition, then how to explain it in each AI tone.

• Orbit

  • Technical: one goal or idea, tracked in its own space. A small team of AI worker roles carries out the executable plan inside it, guided by the user's interactions.
  • Explaining to the user:
    • Casual: "This is your space for one big idea — everything about it lives here."
    • Professional: "A dedicated workspace for a single goal, containing everything needed to execute it."
    • Creative: "Your own little universe, built around one idea you're bringing to life."

• Task

  • Technical: one piece of work within an Orbit, assigned to one AI worker role. Opening a task starts a chat with that worker; the worker produces an Artifact. Once a task has a real artifact, it gets a stable reference number (shown on its card, e.g. "002") — use this to refer back to a specific task/artifact unambiguously when the user or the conversation needs to distinguish between several.
  • Explaining to the user:
    • Casual: "This is one piece of the work — you chat with the AI here, and it makes you something real."
    • Professional: "A discrete unit of work assigned to a specific AI worker, producing a deliverable artifact."
    • Creative: "One spark inside the orbit — talk it through, and watch it become something real."

• Artifact

  • Technical: the actual deliverable a worker produces when a task is executed — copy, a design brief, a strategy doc, etc. The user reviews it and accepts it or asks for revisions.
  • Explaining to the user:
    • Casual: "This is the actual thing the AI made for you — read it, and tell it if you want changes."
    • Professional: "The deliverable produced by the assigned AI worker, pending your review and approval."
    • Creative: "The real thing, made real — yours to shape further or make official."

• Versions

  • Technical: each time a worker generates output it's a new version (v1, v2, v3...). The user decides which version to accept; only an accepted Artifact commits to the Orbit.
  • Explaining to the user:
    • Casual: "Every time it makes something new, that's a new version — nothing counts until you pick one."
    • Professional: "Each generation is saved as a distinct version; only your accepted version becomes final."
    • Creative: "Every attempt gets its own chapter — you choose which one becomes the story."

• Follow-up suggestions

  • Technical: accepting an artifact also generates follow-up task suggestions the user can approve — shaping what the orbit does next, based on everything already accepted. For Launch orbits, these suggestions also lean toward what the orbit's current phase (see below) is actually for — a soft bias, not a hard filter, so a genuinely good idea for a later phase can still surface.
  • Explaining to the user:
    • Casual: "Once you approve something, it'll suggest what to do next — you decide if you want to take it."
    • Professional: "Acceptance triggers suggested next steps, which you may approve to continue the orbit's plan."
    • Creative: "Every accepted step lights the way to the next one — yours to follow or leave behind."

• Your orbit content — the goals, plans, and artifacts you're actually building — stays saved in your own Google Sheet + Drive, always yours.

Every orbit is one of three types, chosen at creation and changeable later. Only Launch orbits get a phase roadmap (see Phasing below) — Pulse and Cycle orbits run purely on Momentum (task activity), with no phase concept at all.

• Launch 🚀 — "A mission with a finish line. Ship it." A project with a real endpoint; gets the phase roadmap.
• Pulse 💚 — "An ongoing mission. Runs until you say stop." Continuous, no finish line, no phases.
• Cycle 🔄 — "Continuous development. Always evolving." Also continuous, no phases.

• Foundation

  • Technical: every Launch orbit starts here, before any phase roadmap exists. Not tied to any specific category of orbit — it's the universal first phase for all of them.
  • Explaining to the user:
    • Casual: "This is the ground floor — get enough real context down here, and it'll offer to map out the rest."
    • Professional: "The starting phase for every Launch orbit, before a phase roadmap has been generated."
    • Creative: "The first spark before the whole constellation takes shape."

• Phase roadmap

  • Technical: once a Launch orbit has enough real context (tasks, accepted artifacts), the AI offers to generate the rest of the phase sequence suited to what's actually being built. Generating the roadmap does not by itself leave Foundation — advancing to the next phase is always a separate, explicit step the user takes.
  • Explaining to the user:
    • Casual: "Once there's enough to go on, it'll offer to map out the rest of the journey — you still choose when to actually move into the next step."
    • Professional: "A generated sequence of phases suited to the orbit's goal, offered once sufficient context exists. Advancing through it remains a separate, user-driven action."
    • Creative: "The map unfolds once the story's got enough weight behind it — but you're still the one who decides when to turn the page."

• Advancing / reverting phases

  • Technical: the user can attempt to advance into the next phase at any time; the AI reports whether it looks ready (with reasoning either way) but never blocks the move — the user always gets both an Advance/Proceed option and a Not-Ready/Cancel option. The user can also revert back into any already-completed phase to keep developing it further; nothing accepted is ever lost by doing so (only later phases reset to upcoming, to be earned back into again).
  • Explaining to the user:
    • Casual: "You can push forward whenever you want to check if it's ready — or step back into an earlier phase if you want to keep building there. Either way, nothing you've already approved goes away."
    • Professional: "Phase advancement is always user-initiated; the AI's readiness assessment is informational, not a gate. Reverting to an earlier phase is also available and never affects previously accepted work."
    • Creative: "Move forward when you're ready to leap, or circle back to build something out further — the story you've already told stays exactly as you told it, either way."

Same two-layer format: technical definition, then how to explain it in each AI tone. (This is about how to describe the worker's role to the user — a separate thing from that worker's own internal personality/ voice, which is fixed per-role regardless of the user's chosen tone.)

• BrainstormAI

  • Technical: Strategist role — ideation & strategy tasks.
  • Explaining to the user:
    • Casual: "This is your ideas person — good for figuring out direction and strategy."
    • Professional: "The strategist role, focused on ideation and strategic planning."
    • Creative: "Your spark generator — where raw ideas take shape into a real plan."

• DesignAI

  • Technical: Designer role — visual systems & UI tasks.
  • Explaining to the user:
    • Casual: "This is your design person — good for how things look and feel."
    • Professional: "The designer role, focused on visual systems and UI."
    • Creative: "Your visual voice — turning ideas into something people can actually see."

• WordSmithAI

  • Technical: Writer role — copy & narrative tasks.
  • Explaining to the user:
    • Casual: "This is your writer — good for anything that needs the right words."
    • Professional: "The writer role, focused on copy and narrative."
    • Creative: "Your wordsmith — shaping the story people will actually read."

• ImageGenAI

  • Technical: Image-generation role. Available now to every account — 3 free generations a day on Starter (10/day on Pro), both resetting at UTC midnight. Running out for the day can be topped up with a one-time ImageGenAI Burst Pack purchase (Momentum Packs page, +10 generations for a rolling 24 hours) — optional, not required to use the worker at all. Most often triggered right after a DesignAI artifact the user accepted suggests generating an image.
  • Explaining to the user:
    • Casual: "This one makes actual images — you get a few free every day, and can top up for a day if you ever need more."
    • Professional: "The image-generation role, included free with a daily generation limit; additional generations are available via a one-time purchase."
    • Creative: "Your visual conjurer — turning a described idea into something you can actually look at, a few times a day for free."

Each worker only sees the context relevant to its own task and the orbit's goal — it does not have access to external websites, databases, company records, or the internet. Anything it produces is a draft for the user to verify, not sourced fact.

Verified against the actual platform code, not approximated:

• BrainstormAI: 💡 orange — #ea580c (solid, no gradient), background #fff7ed
• DesignAI: 🎨 pink/magenta — #db2777 (solid, no gradient), background #fdf2f8
• WordSmithAI: ✍️ blue — #2563eb (solid, no gradient), background #eff6ff
• ImageGenAI: 🖼️ green — #15803d (solid, no gradient), background #f0fdf4

The platform uses DM Sans consistently — confirmed across the landing page, dashboard, task modals, and settings. No secondary heading font; the same family is used throughout.

Verified against the actual platform code:

• Dormant: #64748b (slate gray)
• Pulsing: #a78bfa (light purple)
• Ignited: #f97316 (orange)
• Aligned: #34d399 (emerald green)
• Launched: #f59e0b (amber)
• Beacon: #818cf8 (indigo, cosmetic-only state — see Task States above)
• Cascade: #22d3ee (cyan)

Each state below has two layers: a technical definition (what it actually means in the system) and how it is described to a user (in plain language, matched to the account's chosen AI tone — Casual by default, or Professional or Creative if set).

• Dormant

  • Technical: created, not started. No version has been generated yet.
  • Explaining to the user:
    • Casual: "This one hasn't been touched yet — nothing's happened here."
    • Professional: "This task has not yet been started."
    • Creative: "Still asleep, waiting for its first spark."

• Pulsing

  • Technical: always the state before acceptance, regardless of whether the user chatted with the worker or not. A version has been generated and is awaiting review.
  • Explaining to the user:
    • Casual: "There's a draft ready for you to look at."
    • Professional: "A version has been generated and is awaiting your review."
    • Creative: "Something just came alive — a draft's here, waiting for your eyes."

• Aligned (accepted, and the user chatted with the worker first)

  • Technical: on acceptance, a task becomes one of two parallel outcomes, same level, two different paths to get there. Aligned means the user talked with the worker before accepting.
  • Explaining to the user:
    • Casual: "You talked it through, then approved it — it's locked in."
    • Professional: "Accepted after discussion with the assigned AI worker."
    • Creative: "You two shaped it together, then made it real."

• Ignited (accepted, with no chat beforehand)

  • Technical: the other of the two acceptance outcomes — accepted with no prior chat.
  • Explaining to the user:
    • Casual: "You approved it right away — no back-and-forth needed."
    • Professional: "Accepted without prior discussion."
    • Creative: "Straight to yes — no hesitation, just liftoff."

• Launched

  • Technical: an Aligned or Ignited task that has since generated a confirmed follow-up task — think of it as launching off in a new direction from where it landed.
  • Explaining to the user:
    • Casual: "This one's already kicked off something new."
    • Professional: "This task has generated a confirmed follow-up task."
    • Creative: "This didn't just land — it launched something new."

• Cascade

  • Technical: a task lights up as Cascade the moment any follow-up task it generated has its own artifact accepted (Aligned or Ignited) — the follow-up doesn't need to reach Launched itself, just get accepted. If that follow-up's artifact is later un-accepted (and no other accepted follow-up remains), the original task falls back to Launched.
  • Explaining to the user:
    • Casual: "What you started here is still moving — one of its follow-ups just got accepted too."
    • Professional: "A follow-up task generated by this task has had its own artifact accepted."
    • Creative: "Ripples turning into waves — this is still spreading outward."

• Beacon

  • Technical: shown only on a generated follow-up task still waiting to be confirmed (the translucent "ghost" card on the canvas). Cosmetic only — not filterable, not referenced anywhere else in the platform.
  • Explaining to the user:
    • Casual: "This is just a suggested next step — nothing's official until you say yes."
    • Professional: "A suggested follow-up task awaiting your confirmation."
    • Creative: "A ghost card on the horizon — a hint of what's next, if you want it."

Both kinds are scheduled (daily/weekly/monthly) and both are something the user reviews before it counts — nothing an AutoLoop produces is final until accepted, same as a regular task's artifact. Same two-layer format as Energy states above: technical definition, then how to explain it in each AI tone.

• Momentum Loop

  • Technical: a scheduled recurring job that regenerates a fresh deliverable artifact each cycle, through the same execution pipeline a regular task uses. The user reviews and accepts (or revises) each run before it counts as done.
  • Explaining to the user:
    • Casual: "This one keeps making you a fresh draft on a schedule — you just review it when it's ready."
    • Professional: "A recurring automation that produces a new deliverable on your chosen schedule, pending your review."
    • Creative: "Your own steady drumbeat — a fresh spark delivered right on cue, every cycle."

• Insight Loop

  • Technical: a scheduled recurring job that reads (never produces) — read-only, recurring analysis of the orbit's already-accepted work. What it actually reports is shaped by the directions the user set for that specific loop: a targeted recommendation, an answer to a standing question, a pattern to watch for, or a general status check are all equally valid — whatever the loop was set up to track. Only falls back to a generic status briefing (what happened, what changed, what stalled, where momentum is building, what deserves attention next) when the loop has no specific directions set. A directionless loop reports "stable" when nothing material has changed since last time rather than fabricating progress; a directed loop answers fresh every run regardless.
  • Explaining to the user:
    • Casual: "This one watches for whatever you told it to — a specific answer, a pattern, or just a general check-in — and reports back on schedule. You decide what it's looking for."
    • Professional: "A recurring, read-only analysis shaped by the directions you set for it — anything from a targeted recommendation to a general status check."
    • Creative: "Your own quiet observer — trained on exactly what you asked it to watch for, reporting back what it finds."

• Sample preview (shown before a loop's first real run)

  • Technical: an illustrative example run generated automatically when an AutoLoop is first created, before its real schedule has fired — gives the user a concrete sense of what the loop will produce. Once any real run exists it steps aside and never counts toward run history.
  • Explaining to the user:
    • Casual: "This is just a preview of what this loop might make — not the real thing yet."
    • Professional: "An illustrative example generated at creation, shown only until the first real run occurs."
    • Creative: "A sketch of what's coming — the real thing arrives right on schedule."

• Focus Week
  • Technical: a weekly scheduling view for booking time against specific tasks — one-time or recurring sessions, with a reminder before each one. Each session can also export a pre-filled Google Calendar event (title, date/time, notes, link back to the platform) via an "Add to Cal" button — the user can then optionally add Google Meet themselves from Google Calendar's own screen, and gets Google's native push notifications for it even when not logged into the platform or checking email.
  • Explaining to the user:
    • Casual: "This is where you block off real time for your tasks — set a reminder, and it can drop straight into your Google Calendar too."
    • Professional: "A weekly scheduling view for allocating dedicated time to specific tasks, with optional Google Calendar export and reminders."
    • Creative: "Where intention becomes a real appointment — carve out the time, and it follows you into your calendar."

A workspace has one owner. The owner can invite other people into it by email; an invited person accepts and then works inside the owner's workspace alongside them.

• Per-orbit access, chosen at invite. Being in a workspace does not mean seeing everything in it. The owner picks which orbits each person can reach, and can widen or narrow that later. Somebody left with no orbits at all is removed from the workspace rather than left with an empty shell, and they are told when either happens.
• Two roles: editor and viewer. The owner is neither — ownership is its own thing and cannot be transferred from the UI today.
• Orbit Manager is not a third role but a per-orbit permission the owner gives an editor: on those orbits they can create, assign, unassign and cancel work orders and approve new places — a dispatcher. Someone can manage one orbit and simply be crew on another.
• One Drive, the owner's. Everything the workspace produces is saved to the owner's Google Drive and their master Sheet, not split across each member's own Drive. This is why the workspace is described as the owner's: they hold the files.
• A person can belong to several workspaces — their own, plus any they have been invited to — and switches between them in the app. Each is entirely separate: orbits, tasks, artifacts and conversations do not cross between them.

Explaining to the user:

• Casual: "You can bring other people in and pick exactly which orbits they see."
• Professional: "Invite collaborators to your workspace and grant access per orbit."
• Creative: "Your universe can have other people in it — you decide which parts they can visit."

In-app messaging, so people sharing a workspace do not have to leave the product to talk about the work.

• A room per orbit. Everyone with access to that orbit is in its room; nobody else can see it. Access to the orbit is what puts someone in the room, so removing their access removes them from it too.
• Direct messages between any two people who share at least one orbit. Somebody you share nothing with does not appear as an option.
• What a message can carry: text, images, audio and video that play inline, files up to 8MB, GIFs, links with a preview card, and reactions (one per person per message). A message can be edited for 15 minutes after it is sent and not after.
• Translate: a small translate button beside somebody else's message (DMs and orbit threads) and beside an AI worker's reply translates it into the reader's app language, shown under the original. It stays translated until the app is refreshed; translations are not saved anywhere. Each person's translation is theirs alone -- everyone else still sees the original until they tap it themselves. When everyone in the conversation uses the same app language, the button is hidden.
• Unread state shows on the conversation and in the notification bell, and a message can raise a push notification on a device that is not currently looking at the app.

Explaining to the user:

• Casual: "You can message the people you're working with without leaving the app."
• Professional: "Each orbit has a shared room, and direct messages are available between collaborators."
• Creative: "The crew can talk right here, next to the work itself."

A live session for working through one orbit together, started from the Orbit Sync button in the orbit's header. It survives a page reload as a "Rejoin Orbit Sync" button.

• Two modes. Orbit Sync is the default: the session is the orbit, and followers see the view, task and tab the leader opens there. Stepping off the orbit is leaving the meeting, so it asks first -- and asks the LEADER whether to keep leading, which turns on Workspace Sync: the session follows the leader across the app, even into another workspace (their own solo workspace included). Each follower sees each page as themselves wherever their role lets them. When the leader is the owner or that orbit's Orbit Manager, followers who are not in an orbit the leader opens can still see it, view only, for as long as the leader stays there. The same goes for the pages that list across orbits (dashboard, Orbits, Task Command, Focus Week, the Vault, the map, Places, Work Orders): while the owner or a manager leads there, followers see what the leader's lists show -- every orbit for an owner, the managed ones for a manager -- view only, ending a few minutes after the leader moves on. Where a follower still can't see the page (the leader's mail, Focus Week or other personal pages, another workspace, or an orbit led by someone who can't open it to them), they see a card saying where the leader is, or the leader's screen once the leader presses Share screen. The leader sees who can't see the page: their avatar is greyed with a crossed-out eye. Screen sharing works from a computer browser; sharing mail asks the leader to confirm first. "Back to the orbit" ends Workspace Sync.
• The leader's pointer. A leader using a mouse shows their pointer on everyone's screen, landing on the same task card or button even when the windows are different sizes.
• Switching workspaces mid-session reloads only the leader's app: the session reconnects by itself, the leader keeps the lead, and followers see "reconnecting" for a moment. A screen share needs one tap to resume.
• Following is watching, with shared work in the session's orbit. A follower cannot click, tap or scroll the shared screen; trying shows a prompt offering to leave the session. They can type in the shared search boxes, and in the orbit the session started from they can still work with the leader: chat with the AI worker in the open task, add focus notes, and fill in the Add Task / AutoLoop wizard text the leader opened (only the leader creates). They talk through Messages -- the speech-bubble button on the Orbit Sync rail (or Messages in the panel) opens their own orbit rooms and DMs as a personal overlay with a "See through" option. The Orbit Guide is a personal tool, so it is hidden for everyone during a session (except a leader who is sharing their screen). To stop following, they choose Just listen (stay on the call and move around freely, then Follow again any time) or leave the meeting (the eye button on the rail, or the panel).
• Phones can follow and lead. A phone follows with the leader's pointer shown as a soft circle and the hovered control outlined; held upright it is told when the leader is on the Canvas or Hybrid view ("turn your phone sideways"), and sideways it shows it. A phone can also lead -- including Workspace Sync across the workspace -- with its taps shown to everyone as click rings; the one thing it cannot do is share its screen (phone browsers do not offer it), though it can watch someone else's share. Avatars of people on a phone carry a phone badge.
• Passing the lead. The leader can hand over to anyone or step down; stepping down (or the leader leaving) makes the lead open, takes everyone back to the orbit, and anyone in the session can take it. Crew can ask to lead; the owner or an Orbit Manager can take it directly.
• Voice is optional. Join with a mic (you start muted), listen only, or join without sound. Everyone in the session shows as a floating avatar with their mute state and a ring while they speak. The browser asks for the microphone only when someone chooses a mic. The call is for a small group (up to 8 people). On a phone, voice pauses while the app is in the background.
• Getting people in: starting a session can post a Join card in the orbit's chat room; the panel can send the same card to crew members by direct message; and a focus session's start reminder offers Orbit Sync.
• Nothing about a session is saved -- it exists while people are in it.

Explaining to the user:

• Casual: "Hop into Orbit Sync and everyone sees what you open, and you can talk it through."
• Professional: "Orbit Sync is a live, led session on an orbit, with optional voice."
• Creative: "One person drives, the crew rides along, and you can all talk on the way."

Field work: jobs that happen at a customer's or a place's location, or remotely on their behalf.

• Places — saved customers, sites and resources, each with an address on the map. An orbit uses one or more lists of places. Crew can request a new place; the owner or an Orbit Manager approves it.
• Work orders — a task can have any number of work orders, each with its own place (or none, for remote work), a kind (Field visit or Remote), a type template with set steps, written details, and an assignee. Each has a reference like WO-26090002 that customers and crew can quote.
• The assignee's steps — accept or decline; then on route, on site (field visit) or start (remote); notes, materials, a customer signature or a call outcome; pause, reschedule or release; complete (a field visit needs a visit note). Cancelling needs a reason, and a customer request records who asked.
• Who sees what — everyone who can open the orbit sees a work order, where it stands and what has been saved on it; only the assignee updates it; the owner or an Orbit Manager dispatches it. Every order has a timeline: created, each assignment, each step.
• Energy — a Dormant task shows as Pulsing while it has an open work order, and returns to Dormant when none are left. Tasks further along are unchanged apart from a work-order marker on the card.
• My work orders — each person's own list: waiting to accept, in progress, finished in the last 30 days.
• Customer lookup — find a customer and see which orbits they are in, their open work orders and their history before creating a new one.
• History — every order at a place, in any state, each opening its own page. A task with work-order history cannot be deleted while an order is open, and its orders outlive the task.
• Starter and Pro — every part of this works on Starter, with limits on how many work orders can be open at once, how many places can be saved, how many crew members can hold open work, and how far back history reaches. Pro lifts them. Limits stop starting new work, never finishing what is open, and nothing is deleted.

Explaining to the user:

• Casual: "Save your customers on the map, send someone a work order, and watch it move from accepted to done."

• Professional: "Dispatch field and remote work orders against saved customer locations, with set steps, evidence and a full timeline."

• Creative: "Every job gets a pin, a person and a path from 'on my way' to 'done'."

• Task wizard template "Job for a customer": adding a task can start from this template -- pick the customer from the orbit's places lists (or add a new one), describe the job, and the task is created with its work order for that customer. Crew members get a work order request instead, and can ask for a new place (even a new list) that is tied to the task once approved.

Every Momentum Pack is a genuine one-time Stripe purchase, not a subscription — bought on the Momentum Packs page. Each one lifts one specific Starter-plan limit that can genuinely get in the way, and every pack also names a real free alternative that costs nothing but time or a short wait. None are required to use the platform; a user who never buys one loses no functionality, only speed/headroom on that one limit.

Naming: each pack is named for exactly the mechanic it lifts, not a marketing label — Orbit Boost boosts orbit capacity, an AutoLoop Cool Down Skip skips a MomentumLoop's cooldown days for good, an ImageGenAI Burst gives ImageGenAI a temporary burst of extra generations, and a Task Boost boosts task-execution capacity.

• Orbit Boost Pack — $5.99. Unlocks a 5th active orbit slot, permanently, on top of the Starter plan's 4 — the existing 4 orbits are untouched. Stacks if purchased more than once, up to the account's max. Free alternative: archive an existing orbit to free a slot (its tasks and AutoLoops are removed, though the underlying Google Docs stay in Drive — reviving the idea later means rebuilding it by hand, not a one-click restore).

• AutoLoop Cool Down Skip — $0.99, once. MomentumLoops take cooldown days: on Starter a MomentumLoop rests the day after each run (on Pro, one day in seven). An AutoLoop Cool Down Skip skips those cooldown days for one MomentumLoop, so it runs every day it is scheduled. It is the user's to keep: it can be moved to another MomentumLoop in any orbit from that loop's Settings, and it survives the loop or orbit being deleted and a workspace reset (lost only if the account is deleted). One pass per MomentumLoop; to have several MomentumLoops running every day, buy one for each. A pass that is not in use is offered before a new purchase. Free alternative: leave a gap between the days picked, and every one of them runs.

• ImageGenAI Burst Pack — $1.99. Adds 10 more ImageGenAI generations for a rolling 24 hours, stacking on top of the account's normal daily baseline (3/day Starter, 10/day Pro). Unused generations expire rather than rolling over. Free alternative: every account already gets its free daily baseline at no cost — this pack is only for a day that needs more than that.

• Task Boost Pack — $2.99 ($0.99 for an account's first one). Unlocks 10 more task executions for a rolling 24 hours, on top of the platform-wide 25/day execution cap shared across BrainstormAI, DesignAI, and WordSmithAI. Unused executions expire rather than rolling over. Free alternative: wait for the shared cap to reset at midnight UTC — everything already generated stays available to review and accept in the meantime.

Brief, conceptual — what to expect on each dashboard page, not a pixel mirror of the current UI (layout/wording can shift; what each page is for doesn't). Useful as a talking point when discussing the platform with someone who isn't chatting with its own in-app AI directly.

• Dashboard (home) — the landing page after login: a quick "where do things stand" snapshot (active orbit count, login streak, a small weekly stats summary) plus the same orbit cards as the Orbits page, for fast access into one. Not a management page.
• Orbits (overview) — every live orbit as a card ("pad"), plus an archive section below with relaunch/delete. Where orbit-level housekeeping happens — archiving, relaunching, buying more orbit capacity. Also shows the Badges carousel, an in-app achievement system.
• An orbit's own workspace — one orbit's full working space: its tasks on a visual canvas (or list/hybrid view), each task's chat thread and artifact, its AutoLoops, and — for Launch-type orbits — a phase roadmap tracking progress toward a finish line.
• Tasks (Task Command Center) — a cross-orbit list of every task the user has, in one place, with search and energy/orbit filters — for finding and jumping to a task, not working on it directly.
• Focus Week — a weekly scheduling view for booking real time against specific tasks, with optional Google Calendar export and reminders.
• Orbits Map — the map of saved places with work-order pins and counts; open a place for its details, work orders and history, or look up a customer.
• My work orders — the signed-in person's own work orders, and each work order's own page: its steps, details, assignee, progress, place and timeline.
• Customers & Places (workspace owner) — customers and their locations live in LISTS; every location belongs to one. Each list has its own name (renamable), the orbits that use it (one list can serve several orbits), and a grid of its locations, editable row by row. "Add to this list" offers One location (a search), Several locations (describe them; real ones come back from the map) or From a file (a spreadsheet, Word doc, PDF or text file of customers — a sheet is read by its column headers, other files by AI); tick which to keep. "New list" can start blank, from a file, or from the map. Adding a location from inside an orbit asks which of that orbit's lists it joins.
• The Vault — a read-only, searchable archive of every artifact the user has accepted across every orbit — the permanent home for finished work, not a place work gets created or edited.
• Momentum Packs — one-time paid boosts for a Starter-plan limit that's genuinely in the way right now (more orbit room, skip a Momentum Loop's cooldown day, extra ImageGenAI generations, extra task executions) — never required, every pack names its own free alternative. See the Momentum Packs section below for exact pricing.
• Settings — profile and avatar, "Orbit Guide Voice Tone" (see AI Tone below for what this actually controls), interface language, the connected Google account (and a link to the user's real underlying Google Sheet/Drive data), plan status, and account actions.

VisionAI runs on a tiered, multi-provider architecture across Groq, Google Gemini, Mistral, and OpenRouter — this is public information, already published on the platform's own "Under the Hood" page, not a secret. Lightweight conversation is routed to fast, efficient models; higher- capacity models are reserved for generating actual deliverables, and a chat response gets automatically upgraded to a heavier model mid- conversation if the accumulated context genuinely calls for deeper reasoning. Every request rotates across providers and keys for speed, and if any model or provider is unavailable, execution automatically fails over to the next one — the platform keeps moving without interruption. Task routing to a specific worker (BrainstormAI/DesignAI/WordSmithAI) is based on task/domain fit, not a manual dropdown choice.

Same three options (Casual / Professional / Creative) in both places, but they're genuinely separate settings that can be set differently, not one setting with a fallback/override relationship:

• Account Settings → "Orbit Guide Voice Tone" — set on the main Settings page, saved to the account. The Settings UI itself frames this as governing only Orbit Guide's chat voice ("How your Orbit Guide communicates with you"), but it also silently sets the wording style used when generating Momentum/Insight Loop (AutoLoop) output — that second effect isn't mentioned in the Settings UI copy. Defaults to Casual.
• Per-orbit tone — set from inside a specific orbit's own settings, not from the main Settings page. Governs the wording style for that orbit's task chat, artifact generation/revision, and follow-up suggestions. Independent of the account-level control above — it does not inherit or default from it, so a user can genuinely have Orbit Guide set to Professional while a specific orbit's task chat is set to Creative. Also defaults to Casual.

What "tone" actually changes: real wording-style guidance injected into the AI's instructions (Casual and Creative each carry concrete before/after phrasing examples; Professional gets no extra instruction since formal register is already the model's natural default) — it changes how something is said, never what the platform can do or which features are available.

Three plans exist: Starter (the free plan), Pro, and Enterprise. Always call the free plan "Starter" ("the Starter plan"), never "the Free plan". Plain factual description per plan — no tone variants here, since these are fixed facts about pricing/scope, not something to phrase differently by chat tone.

• Starter plan (free) — 4 active Orbits (extendable to 6 via Orbit Boost Pack purchases, see Momentum Packs above); all 3 core AI workers; 3 free ImageGenAI generations a day (Burst Pack purchase available for more); 2 MomentumLoops + 1 InsightLoop in EACH orbit (per orbit, not shared across the account -- owner, 2026-10-14); everything saved to the user's own Google Drive; no expiration on existing orbits. Free permanently, no forced upgrade.

• Pro plan — still in development, opt-in only, join-the-waitlist stage; no pricing announced yet. Confirmed scope so far: raises the ImageGenAI free daily baseline to 10 generations a day (vs. 3 on Starter), 8 active Orbits (vs. 4 on Starter — also stacks with Orbit Boost Pack purchases up to +2, same as Starter, reaching 10 max), and Momentum Loops run 6 days before cooldown instead of Starter's 2, and double the loop counts — 4 MomentumLoops + 2 InsightLoops in each orbit, against 2+1 on Starter (LOOP_CAP_EXECUTION_PRO/LOOP_CAP_INSIGHT_PRO in app/lib/autoloop/caps.ts). This file previously said Pro did NOT raise those counts, which contradicted both the pricing card and llms.txt and was wrong against the code.

• Enterprise plan — no self-serve pricing or signup; the user submits a Contact Us request and gets a custom quote for what they actually want implemented, rather than a fixed package. Confirmed scope: everything in Pro, plus custom orbit limits, custom-tailored AI workflows, dedicated onboarding & support, and custom integrations on request.

Maintained manually. Update this file whenever something described above changes in the real product — a stale reference file is worse than none.