Contents
Every app now has its own agent: one in your docs, one in your inbox, one in your CRM. Each behind its own button, each blind outside its own app, each meeting you like a stranger. Using them means going to them.
Vestra works the other way. Press one key over whatever you are doing and Bash is there: already seeing your screen, already knowing your work, able to act across every tool you use, on its own computer or on yours. Say it or type it, then get back to work. It comes back done.
Underneath that keystroke is a workspace staffed by AI agents that take outcomes, not instructions. They plan the steps, run the tools, hand back finished work, and learn about your world. Delegation, no longer rationed.
It becomes someone you call.
Here is what that means on an ordinary Monday.
- Last week’s invoices went out, reminders queued behind them
- Yesterday’s orders and refunds reconciled, books clean
- The proposal for the new lead is drafted, ready for your read
- The proposal quotes a 14% discount, steeper than your usual. Send it?
- One refund is over your auto-approve limit. Yes or no?
Refunds jumped 3× since Tuesday. Cause traced, fix drafted for your sign-off.
You didn’t do any of this work. You delegated it.
That is the whole idea. Vestra is an autonomous workspace where humans and AI agents collaborate to move business work forward. Instead of starting with a blank chat window, you start with an outcome: the invoices, a proposal, a client report, a campaign. The work comes back done.
If you have tried AI tools before, you may be braced for the catch. Fair. Most of them are shaped like a line: you prompt, it answers, and everything it learned about your business evaporates on the way out. Vestra is shaped like a loop. Two engines make it turn.
Look back at that Monday briefing. Every piece of the workspace was already in it, and every piece sits somewhere on the loop:
The doer. Plans the task, operates tools, browses the web, writes files and code, and carries every request from asked to done.
The rememberer. Quietly builds a living brain of your business (people, projects, deals, decisions, history) from your work and your connected tools.
Bash acts; Nami learns from every action; the next task starts smarter than the last. That loop is why Vestra compounds: the workspace you have on day 100 is meaningfully more capable than the one you signed up with on day 1.
That is the whole shape: work goes into the loop; finished work and memory come back out. Everyone starts the same way, by handing over one task in plain words. The rest introduces itself as your work calls for it.
The rest of this handbook walks the loop one piece at a time: what each piece does, where it pauses to ask you, what it costs, and what it cannot do yet. Start with the engines.
One agent does the work. One agent remembers everything. The loop between them is the product.
Bash: The agent that gets things done
Bash is the doer. Give it an outcome in plain words, like “get this month’s invoices out” or “reconcile this week’s payouts”, and it writes a plan, picks its own tools, and carries the work all the way to done: browsing the web, pulling data from your apps, writing files and code, producing the finished deliverable. And it is a lead, not a solo act: for specialist work Bash calls in specialist agents (a researcher, an analyst, a reviewer) and runs them in parallel under its plan. You watch every step, and the unit of work stops being a message. It becomes a finished task.
Delegation only works if someone carries the work to completion. Bash does, choosing and operating the right tools along the way, so you hand over whole outcomes instead of guiding an AI through every step. And because each task runs in its own session, you can hand over ten at once and close your laptop.
- Hit New Task, describe the outcome in normal words.
- Bash writes a short plan, then works through it while you watch.
- It pauses only when it genuinely needs you (a login, a permission, a plan to approve), then resumes in place.
- Steer mid-task with a message, or stop it anytime.
- Get the month’s invoices out, with reminders queued
- Reconcile a week of payouts and flag what doesn’t match
- Turn a supplier’s new price list into updated listings
- Generate a product video and build the marketing campaign around it
- Turn a messy spreadsheet into a clean, labelled dashboard
- Research what competitors charge and hand back a comparison
Bash plans before it acts, shows each step, and pauses instead of guessing on anything risky. If it cannot finish, it says exactly where and why. Nothing fails silently. Long jobs run in the cloud, so you can close your laptop.
Nami: The agent that builds the brain
Nami never runs your tasks. It does something quieter and, over time, more valuable: it watches everything: every task Bash finishes, every document, every decision, plus the Slack threads, emails, CRM records, files, and calendars in the tools you connect. All of it weaves into one living, structured understanding of your business. The people. The projects. The deals. What was promised, and to whom.
Context is the fuel for autonomy. The richer the brain, the more your agents do without being told: they already know who the customer is, what was promised, where the file lives, and what happened last time. And because Nami keeps watching your connected tools, that understanding stays live, not a one-time snapshot you have to maintain.
- You never operate Nami; it learns automatically as you work.
- Connect a tool once and Nami backfills from its history (past threads, deals, and documents), then keeps learning as new work happens.
- Ask it anything: “what did we promise Brightside on the kickoff call?”
- It reconciles as things change, merging “Josh” and “Josh A.” into one person, updating a title after a promotion.
- Gives every new task your whole company’s context, no setup
- Answers questions across Slack, Gmail, and your CRM at once
- Briefs you on any person, deal, or project in seconds
- Keeps facts current as emails, deals, and docs change
- Surfaces commitments and follow-ups before they slip
- Saves you from ever re-explaining your business
You never set Nami up or maintain it; it builds itself as you work. It is unique to your workspace and private to you. Because it compounds, it is most valuable on day 100, not day 1. But day 1 is not zero: Nami seeds the brain from your connected tools’ history, so it starts out already knowing who your customers are.
The Autonomy Loop
This is the part that makes Vestra autonomous, not just helpful. Bash executes a task and, in doing so, creates new context: what was done, what was decided, what was learned. Nami captures and structures it. The next task starts from that richer brain, so Bash plans better, asks fewer questions, and handles more on its own. Every pass tightens the loop.
The loop does not always have to start with you. As Nami’s understanding deepens, it begins to notice what needs doing on its own: a renewal coming due, an email that needs a reply, a number that looks off. Increasingly, Nami creates the task and hands it to Bash before you ask; you move from assigning work to reviewing outcomes Nami proposed. The work starts finding you.
- Act → Bash gets the outcome done across your tools.
- Learn → Nami turns what happened into lasting context.
- Improve → the next task runs smarter and more autonomously.
Work that used to need your guidance becomes work Bash does alone. Over time the system shifts from helping your team work to working alongside them, and increasingly ahead of them. That “ahead of them” is not a mood; it is machinery, and it has a name.
The Anticipatory Engine: Work that starts before you ask
Everything so far begins when you summon it. The anticipatory engine is the loop running in the other direction: Nami watches the same signals you would (the calendar, the inbox, the numbers, the promises people made) and starts the work before you call. Not a feed of alerts. Finished thinking, delivered at the moment it is useful.
Anticipation spends the scarcest thing you have: attention. So the engine keeps score. Every brief you skim, every prep you open, every proposal you approve or dismiss teaches Nami what deserves your time, and every miss costs it standing. It would rather stay silent than be noise: if nothing matters, it says nothing.
- 8:54 am · the briefing that opened this handbook: finished work, two decisions, one catch.
- 9:26 am · four minutes before the client call: the prep pack, including the discount you promised in March.
- 2:10 pm · “Refunds jumped 3×. Cause traced, fix drafted. Ship it?” One tap.
Every proposal waits inside the autonomy you have granted: anticipation changes when work starts, never who approves it. And when you catch yourself approving the same proposal every week, that is the engine telling you it is ready to become a duty. The next chapter climbs that ladder rung by rung.
Three rungs of delegation: a task you hand over, a routine that repeats, a duty Bash owns outright. Each step up hands Bash more of the thinking, from the doing to the deciding.
Tasks: The work you hand over
A task is the atom the whole workspace is built from: one outcome handed over, in plain words, run to completion in its own session, delivered back finished. Asked, worked, delivered, done. And because every task is a real object, with a plan, a live session, a history, and a deliverable, you can watch it, steer it, replay it, or promote it up the ladder.
A task is delegation made concrete. Each one runs on its own, in parallel, in the cloud, so handing off ten outcomes costs you no more attention than handing off one. You stop sending messages and start handing over results you expect back.
- Describe the outcome; watch the plan, then the steps, run live.
- Start as many as you like; each is independent.
- Steer mid-run with a message; pause or stop anytime.
- Re-run a past task on fresh data, or pick it up from history.
- Kick one off from your phone, finish it on your laptop.
- Run ten at once, each in its own session
- Steer one mid-run without starting over
- Stop a task and resume it where it left off
- Re-run a past task on this week’s numbers
- Kick one off from your phone, finish on your laptop
- Promote one you keep repeating into a routine or duty
The tasks you repeat on a schedule become Routines. A whole standing outcome becomes a Duty. Every duty is, underneath, just the tasks it decides to run on your behalf: the same atom, now watched and held.
Routines: Work that runs itself
A routine is a task with the starting handled. It can be scheduled (run once at a set time), recurring (every Monday at 9am), or reactive (fire when something happens: a support email lands, a bug is reported, a deal changes stage). You set the trigger once; from then on the work starts the moment it should, with nobody remembering to kick it off.
Businesses run on three kinds of repeatable work: the scheduled, the recurring, and the reactive. Routines cover all three, turning Bash from something you ask into something that responds on its own. The Monday report writes itself, the one you used to write at 9pm on Sunday. The new lead gets a CRM record before you have seen the email.
- Describe the task once.
- Choose a trigger: a schedule, or an event in a connected tool.
- Results land in Home and your chosen channel.
- Pause, edit, or switch any routine off anytime.
- A revenue digest in Slack every Monday morning
- Daily payout reconciliation, untouched by humans
- New bug → triaged, labelled, first reply drafted
- New lead → CRM record created, response sent
- Invoice arrives → filed and logged
- Usage limit hit → team alerted with context
In a routine, the thinking is yours: you write the trigger and the task, and Bash executes. When you want Bash to do the thinking too, deciding what to watch and inventing the work itself, that is a duty, one rung up.
Duties: The work Bash holds
A duty is a whole responsibility, handed over: “never let a renewal slip.” “Own our revenue reporting.” “Keep support triaged.” With a routine, you did the thinking: you named the trigger, you named the task. With a duty, Bash takes over the thinking. It decides what is worth watching, creates its own tasks and routines, rebuilds its plan as the real world changes, and reports back. You delegate the loop, not the steps.
Until now, AI automated the doing while you kept the rest: noticing what needs attention, deciding what to do, starting the work, following up. A duty hands Bash that whole loop: from completing tasks to holding responsibilities. And because every duty carries a scope, an authority limit, and a ledger, you delegate the way a business actually does: an outcome, to someone accountable.
- State the outcome and the limits: “Own our renewals. Propose before anything is sent.”
- Bash breaks it down itself: what to watch, which routines and tasks to run.
- It re-plans as reality shifts: a new account, a price change, a telling email.
- It acts within its limits, or drops a proposal in your Needs you lane.
- It earns autonomy: approve a proposal a few times and Bash offers to own that step.
- Open the ledger anytime: every check, decision, and task it invented, on the record.
- Never let a renewal slip
- Own the weekly revenue report, end to end
- Keep the support inbox triaged and answered
- Watch competitors and flag moves that matter
- Chase invoices before they age
- Keep the books reconciled, always
The cleanest test: who writes the trigger and the task? In a routine, you do. In a duty, Bash does, and rewrites them as the situation changes. A routine automates a step you designed. A duty delegates the designing.
What your agents work with: the expertise they carry, the tools they reach, the memory they draw on, the machine they run on, and the software they can build you.
Abilities: Teach it a whole new job
An Ability gives your agents a new specialty: not a single trick, but everything needed to do a real job well: skills, specialist agents, tool connections, hooks and guardrails, custom interfaces, and quality checks, bundled and installable in a click.
The Ability shaped to how your business actually works
Ready-made Abilities give your agents instant range. The real shift is a custom Ability, built around your own way of working: your playbooks, your data, your tools, your standards, encoded once so every agent runs the job the way your company does. Most teams have theirs shaped during onboarding, so Vestra fits your business from day one instead of you bending to a generic tool.
Off-the-shelf AI is a capable generalist. An Ability turns your agents into specialists that do a real job, to a real standard: it bundles the skills, specialist agents, tool connections, guardrails, and quality checks a job actually needs, so the work holds up instead of merely looking right. You grow what Bash can do the way a team grows: install an Ability and your agents gain a whole profession in a click, then stack several so one task can move across sales, accounting, and content at once.
- Browse the Abilities marketplace; install in a click; agents gain the skills and connectors instantly.
- Use first-party and partner Abilities (like Content, powered by Higgsfield) or community ones.
- Get a custom Ability built for you, around your playbooks and data.
- Stack several; one agent switches between jobs as a task demands.
- Install Content and brief a full launch campaign
- Install Sales and run pipeline, follow-ups, and forecasts
- Install Accounting and take receipts through to quarter-end
- Have a custom Ability shaped to your exact workflow
- Combine abilities so one task spans several jobs
- Share every installed ability with your whole Space
An Ability is more than a skill: it brings its own agents (new specialists for Bash to call on), MCPs, and evals. Ready-made abilities give you instant range; custom abilities are how Vestra becomes truly yours. And the catalog keeps growing on its own: new and sharpened Abilities arrive as curated drops — nothing for you to update.
Connectors: Where the real work lives
Real work happens inside your tools, so Vestra plugs into them. Connectors link your agents to Stripe, your CRM, Drive, databases, Slack, email, and 2,000+ more: reading the real numbers, updating the real records, sending from the real account. Always with your permission.
Without connectors, an AI hands you instructions to carry out yourself. With them, Bash works where your business actually lives: the gap between “here’s a draft for you to send” and “sent, from your account, logged in the CRM.”
- Connect a tool once with a secure, revocable connection.
- Agents use it automatically on any task that needs it.
- Abilities bring their own connectors; installing one wires up its tools too.
- Pull live revenue and payouts from Stripe
- Update a deal or contact in your CRM
- Read and write files in Google Drive
- Query your own database for the real numbers
- Post a message or file into Slack
- Create and update calendar events
A connector is how Vestra reaches your tools; a channel is how you reach Vestra. Connections are secure and revocable, and your credentials are never exposed to the model (see Security & Privacy).
Knowledge: What your agents draw on
Knowledge is the living memory every task starts from: the company brain Nami builds, seen as a layer you can use. It holds your documents and the facts of your business, but also your preferences (how you like things done), your rules, your connected tools, past episodes (what happened on earlier tasks), and the relationships between people, projects, and deals.
Agents are only as good as what they know. Knowledge is why yours act like they have worked at your company for years: they follow your preferences, respect your rules, and remember what happened last time. The deeper it gets, the more on-brand and autonomous the work becomes.
- Drop in documents; agents save their outputs back.
- Set rules in plain language: “always CC our bookkeeper on invoices.”
- Browse and shape what Vestra has learned.
- Ask questions across all of it at once.
Knowledge and Nami are the same memory seen from two sides: Nami writes and curates it in the background, you read and shape it here, and every agent draws on it. See Wiki & Vault for the fully-open version.
Computer: Shared between you and your agent
Every task runs on a real remote computer (a genuine filesystem, a live browser, a full terminal) that you and your agent share. It lives in the cloud, persists across sessions, and gives the agent a real place to keep files and do work. Real work needs a real machine.
Files to keep, a browser to log into sites and click around, a terminal to actually run things. And because the machine is shared, it is collaboration, not a black box: watch the agent work live, grab the keyboard for one tricky step, hand it back.
- Open any task → the Computer panel: Files, Browser, Terminal.
- Watch the agent browse in live pixels; click in and co-drive the same session, no ceremony.
- Every file it creates lands under My files, kept across sessions.
- Upload sources; download finished deliverables.
The machine lives in the cloud, not on your laptop: nothing runs on your device, and tasks continue when you are offline. You can grab control at any moment, and your files stay private by default.
Apps: The tools your business is missing
Every business has gaps a spreadsheet or generic SaaS only half-fills: an internal dashboard, an intake form, a small workflow tool. Describe the one you need and Bash builds and deploys it: a live app backed by a real database, wired to your Connectors and Knowledge, running on Vestra’s infrastructure or your own. No code, no backlog.
Bash does not just do the task; it can build the tool that does the task. The dashboard your team has wanted for a year runs this afternoon, shaped to how you actually work, instead of sitting in an engineering backlog.
- Describe the app or dashboard in plain words.
- Bash builds it with a real database, wired to your data.
- Deploy for the whole team or just for you.
- Ask for changes anytime; the live app updates.
- A live revenue or pipeline dashboard
- A client intake form that updates the CRM itself
- A customer health board that updates itself
- A job tracker for the crew in the field
- A personal app just for you: a habit tracker, expense log, or personal CRM
- A report anyone on the team can open and filter
Apps come with a real database, not a static view, and deploy on Vestra’s infrastructure or your own. They stay wired to your Connectors and Knowledge, so they run on live data.
What you actually touch each day: your command center, your team’s shared space, your knowledge in the open, and every app you already use.
Home: The Cockpit
Home is the screen from the top of this tour. It opens with a short briefing of what happened while you were away, then organizes everything by what you need to do about it: what needs your decision, what Nami is proposing, what is running now, what is coming up. Tasks you started and tasks Nami started, side by side.
Two decisions over coffee. The rest is already in motion.
When agents run work in parallel, and increasingly start work on their own, a flat list of chats stops being enough. Home is the view a good chief of staff would hand you each morning: here is what got done, here is what needs you, here is what I would do next. Organized around your decisions, not a feed of activity.
- Start the day with the briefing: what finished, what is waiting.
- Clear the Needs you lane: approve, answer, sign off.
- Failed or stuck tasks land there too, with the reason and the work so far. Steer, retry, or take over.
- Review what Nami proposes: run it, edit it, or dismiss it.
- Peek into any live task to watch or steer it mid-flight.
As Nami learns your business, Home gets more proactive: it shifts from a list you fill to a briefing that fills itself. And as you hand over duties, the briefing becomes their report: which are healthy, which caught something, which need your call. You stay in charge throughout: nothing acts beyond the autonomy you have granted.
Spaces: One workspace, Your whole team
Real work is rarely done alone, so Vestra isn’t a private chat on one person’s screen. A Space is a shared, multiplayer workspace where your team works alongside the same agents, on the same company brain, with the same tasks, files, and abilities. See who is working on what. Hand tasks back and forth. Let the agents serve everyone at once.
The knowledge one person builds and the abilities a colleague installs are there for everyone, so the more people work in a Space, the smarter it gets for all of them. Compounding, multiplied by headcount. This is how Vestra scales from one operator to a whole company.
- Create a Space per team, department, project, or client.
- Invite people; everyone delegates to and steers the same agents.
- See who is on each task; hand work off in a click.
- Keep a private Space just for yourself.
Agents, context, and abilities are shared inside a Space, while permissions keep sensitive work scoped to the right people. The bigger the team, the more the shared brain compounds.
Wiki & Vault: Navigate your knowledge
When agents act on what they know, you need to see that knowledge, trust it, and fix it when it is wrong. So the memory Nami builds is never a black box: the Wiki is a browsable, cross-linked knowledge base, with a page for every person, project, customer, and decision, each one editable. The Vault is the same knowledge as a real Obsidian vault: plain Markdown files with backlinks and a graph, yours to open, edit offline, and keep.
Wiki
Browse everything Vestra knows (company, people, projects, customers, decisions) as linked, editable pages. Correct a page once and every future task uses your correction.
Hosted · linked · always currentVault
The same knowledge as plain Markdown with [[backlinks]] and a graph view. Open it in Obsidian, edit offline, use any plugin; it syncs back. Portable, and yours to keep.
Obsidian · Markdown · no lock-inNothing is hidden, everything is editable, and a correction you make becomes the version every future task uses. It is also yours to take: because the Vault is plain Markdown, your company’s memory is never locked inside Vestra.
- Read the page on any customer, project, or decision.
- Edit or delete any fact; it sticks from the next task on.
- Open the Vault in Obsidian: graph, backlinks, plugins, offline edits.
- Everything stays in sync: Nami writes, you refine, agents read.
Delete anything you do not want remembered. Export everything as plain Markdown, anytime. Your knowledge is never locked in.
Channels: In every app you already use
The best tool is the one that is already open. Delegate from the web app, WhatsApp, Telegram, Slack, email, or a phone call. Message Bash like you would message a colleague, and the result comes back in the same place. Same workspace, same brain, every channel.
No new habit to learn, no tab to remember. Vestra becomes part of the workflow you already have: forward an email and say “handle this”; text it from the road; kick off a task from a Slack thread without leaving it.
- Connect a channel once from settings.
- Message Bash like a colleague: “draft a reply to this and book a follow-up.”
- Context never resets when you switch channels.
A channel is how you reach Vestra; a connector is how Vestra reaches your tools. Same words, opposite directions, and you only set each up once.
Omnipresence: One agent around your whole day
Channels answer “how do I reach it?” Omnipresence answers the better question: where is it when you need it? Everywhere your work happens. On your desktop a keystroke away, on your phone before you sit down, in your inbox, in your meetings. Not an app you open: a presence that follows the day, carrying the same brain the whole way.
Being everywhere is only a gift if it knows when to stay quiet. Vestra does not narrate your day or ping you to prove it is working. It surfaces when something deserves you (a decision, a catch, the brief you expect) and is silent the rest of the time. Presence means available, not audible.
- The morning brief finds you wherever you are; no tab required.
- At your desk, Vestra sits one keystroke away over whatever you are doing — and it sees what you see: summon it over any screen and it already has the context. Speak or type; get back to work.
- Dictate a reply and Vestra types it into whatever app you are in, polished, in your voice.
- Grant it an app — once, and only that app — and it has hands: clicking, typing, and scrolling on your machine, under the same autonomy dial as everything else.
- Walk into a meeting and the prep is already there; walk out and the follow-ups are captured.
- Away from the desk, any channel is the same agent. Nothing resets.
Channels are the doors; omnipresence is the posture. The web app and today’s channels are live, and the desktop presence is real today in private beta: a quiet bar one keystroke away that sees your screen, takes your voice, and — with your grant — lends a hand in your apps. Meeting surfaces are rolling out on the board. They all serve one promise: you stop going to the workspace, because the workspace travels with you.
You hold every key: you set the limits, your data stays yours, and nothing locks you in: not the infrastructure, not the AI model.
Permissions & Approvals: Stay in Control
You decide what agents may do on their own and what needs your sign-off, per workspace, per task, per duty. Bash pauses for anything sensitive (a send, a payment, a login, a plan worth approving) and asks a clear question instead of guessing. Autonomy is a dial, not a switch.
Approvals give you the speed of delegation with the safety of a final say. And the dial moves as trust builds: approve the same kind of proposal a few times and Bash offers to own that step. Autonomy is earned, visibly, never assumed.
- Pick a permission mode, from “ask me before each action” to “run freely.”
- Approve or tweak a plan before the agent commits to it.
- Grant a login or one-time permission inline; the task continues in place.
- Tighten the dial for sensitive work, loosen it for routine tasks.
Nothing important happens behind your back: sends, payments, and anything irreversible always pause at the level of autonomy you have set, on every task, routine, and duty.
Security & Privacy: Keep it safe
You are trusting Vestra with real business data and real logins, so the architecture is built around one rule: your data stays yours and your secrets stay secret. Each workspace is fully isolated. Agents run in a secured sandbox. Your credentials are injected behind the scenes, never exposed to the AI model, and file links are private and time-limited by default.
Connecting your CRM, your books, and your inbox to an agent only makes sense if a compromise of any one layer cannot reach the rest. Isolation per workspace, sandboxed execution, and credential injection are what make “give the agent real access” a reasonable thing to do.
- Connect tools through secure, revocable connections. Never paste keys into chat.
- Keep files private by default; share with time-limited links when you choose.
- Revoke any tool’s access in one place, anytime.
- For strict data residency, run Vestra on your own infrastructure.
The model never sees a password or an API key: credentials are injected at the infrastructure layer, used for the action, and never enter the conversation.
Transparency: Every credit, Accounted to the turn
Vestra meters by credits, and shows you exactly where each one went. Every turn is itemized: the intelligence (the model’s reasoning), the actions (the tools and APIs it called), and the compute (the browser, terminal, and machine it ran on). No black-box bill at month’s end: open any task and read its receipt, down to the turn.
Delegating freely only works if you trust the meter. Per-turn, itemized accounting means no surprises: you see which tasks cost what, and why, then optimize with real numbers, pinning a cheaper model for routine work or trimming a noisy routine, and letting the expensive reasoning run where it earns its keep.
- Open any task to see its receipt, split by intelligence, actions, and compute.
- Roll the ledger up by day, task, routine, or duty.
- Set budgets and alerts; Vestra flags a duty before it burns its allowance.
- Export the breakdown for finance, anytime.
This is the same breakdown behind Plans & Credits: heavier outcomes cost more because they use more thinking, more actions, or more compute, and you can see exactly which.
Intelligence: Any frontier model, Always the latest
The AI landscape changes every week, so Vestra is not tied to one provider. It runs on any frontier model and, by default, picks the best one for each task: a fast model for quick work, a deeper one for hard reasoning. Prefer a particular model, or need to stay with one provider? Pin it.
Every leap in AI flows straight into your workspace: a new model is an upgrade, not a migration. You are never stuck on yesterday’s model, never locked to a single lab, and whatever you swap, your tasks, files, and company brain come with you.
- Let Vestra auto-select the best model per task.
- Pin a specific model when you have a preference.
- Stay on a single provider if your policy requires it.
- Get new models the moment they ship, with no migration.
Channels, infra, intelligence: three swappable layers. The workspace (your tasks, knowledge, abilities) is the part that stays. That is the opposite of how most software locks you in.
Infra: Our cloud, Your cloud, Or on-prem
Where your agents run is a compliance and trust decision, not just a technical one, so it is your choice. Start on Vestra’s secure cloud with nothing to configure. When data must stay inside your environment, point Vestra at your own cloud or machines, and the same agents run there instead.
The things that usually kill a tool’s rollout (security review, data residency, procurement) become a hosting choice instead of a product compromise. Start with zero setup; move when requirements demand it; your tasks and brain come along.
- Start on Vestra’s secure cloud. Nothing to configure.
- Bring your own cloud when data must stay in your account.
- Run on-prem for the strictest, air-gapped environments.
- Move between them as you grow, without losing work.
Your infrastructure choice never changes how Vestra behaves: same agents, same workspace, same company brain, wherever it runs.
Most software peaks the day you buy it, then quietly tapers off. Vestra is good on day one and better every week after: here is what that earns you, and what it costs.
What compounding actually feels like
Every pillar in this tour feeds the same flywheel: tasks create context, context feeds the brain, the brain makes the next task smarter. Here is what that loop feels like from the inside.
You hand over your first task and watch it run. Bash asks more questions today than it will ever ask again; every answer becomes memory. A few tools get connected. The brain starts forming.
The brain holds your clients, suppliers, preferences, and history. Routines run your mornings. Tasks start with context instead of questions, and the work comes back on-brand without being told.
Nami proposes work before you ask. Duties hold whole outcomes: renewals never slip, books stay reconciled. Your briefing reads like a chief of staff wrote it, because functionally, one did.
This is the difference between a tool and a workspace: a tool is the same every time you pick it up. A workspace accumulates. Switching costs do not keep you here; compounding value does.
Plans & Credits
Vestra is a subscription, metered by credits. Every plan includes a monthly credit balance; agents spend credits as they work, so heavier outcomes use more, lighter ones use fewer. Flat monthly pricing keeps it predictable: you delegate freely without watching a meter on every message. For scale: a light task (a draft, a lookup, a quick reconciliation) runs a few credits; a heavy outcome, like the 30-credit board deck earlier, runs tens, not hundreds.
The model is value-aligned: the more work Vestra does for you, the more you get done. Top up anytime without changing plans; move to Business when your team joins, so everyone shares one company brain. Vestra is in private beta: you join with an invite code, and joining comes with 1,000 credits — enough to hand over real work before you pick a plan.
Where We Stand
We ship constantly, and we would rather show you the board than make promises. Here is what is live for your team today, what we are building right now, and what is next. Live means live: if it sits in the first column, your team can use it this morning.
- BashThe agent that does the work, end to end.
- TasksHand over an outcome; it comes back done.
- RoutinesScheduled and triggered work that runs itself.
- AbilitiesTeach Bash how your company does a job.
- Connectors2,000+ apps, reading and writing the real records.
- ComputerA real cloud machine: files, browser, terminal.
- HomeThe morning briefing and the needs-you lane.
- ChannelsTelegram, Gmail, and Slack.
- PermissionsYou set what runs on its own and what needs sign-off.
- SecurityCredentials stay scoped, revocable, and off the model.
- TransparencyEvery credit itemized down to the turn.
- NamiThe company brain that learns from every task.
- DutiesBash holds a whole outcome, proposing before it acts.
- Anticipatory engineBriefs, prep, and proposals that start before you ask.
- Wiki & VaultBrowse and edit everything Vestra knows.
- OmnipresenceDesktop bar, voice, and hands are in beta; meetings next.
- AppsBash builds and deploys the internal tools you are missing.
- SpacesOne shared workspace for your whole team.
- More channelsWhatsApp and phone.
- More models, fasterNew frontier models the week they land.
- Your cloud, or on-premRun the whole workspace on your own infrastructure.
We keep this board current as we ship. The first column is not a preview: it is what your team can lean on right now, today.
Shopify