Juggernaut: a force that, once moving, does not stop
The unstoppable
workforce, always on.
Hand a whole function to a standing AI team and it runs it — without you. It works around the clock, reacts the moment new work lands, remembers everything, and never re-asks you. Not a task you wait on; an operation that runs itself.
Start with one team — a support desk, an ops crew, a research desk — or give it a mission and a CEO builds a whole company of self-running departments. Same dial, hard budget caps either way.
Bring your own model key. One agent or a hundred — hard budget caps either way.
Three steps. Then it's rolling.
If you can write a sentence, you can run a workforce.
Plug in your model and your tools
Add your own model key — Claude, OpenAI, or any compatible endpoint; you pay your provider directly and we never resell tokens. Then connect GitHub, Gmail, Drive, Outlook, or anything with an MCP server. Your team only ever touches what you connect.
Set the team & a hard cap
Choose how many Fast, Smart, and Genius agents it may run, and a dollar cap it can never exceed. Caps, not guesses — then you can walk away.
Hand it a function — and leave
Give it a standing mission and put it on-call. It watches for new work — a fresh email, a ticket, an issue — and handles it on its own, round the clock. You don't re-prompt; you check in.
Fast · your cheapest model
Tireless workers for routine tasks
Smart · a strong mid-tier model
Managers and hard problems
Genius · your most capable model
The director — plans and answers for the job
What you can hand off
Not one-off tasks — whole functions that run themselves. Pick one, or stack several into a company.
Support desk
Watches your inbox, triages every ticket, drafts and sends replies, escalates the hard ones — and learns your product as it goes. You wake to a cleared queue.
Ops & on-call
Watches your alerts and tickets around the clock. Diagnoses, runs the runbook, fixes what it can, and flags what it can't — so nights and weekends are covered.
Research desk
Keep a topic current. It monitors sources, summarizes what's new, and maintains a living brief you can read any morning — without asking.
Engineering team
Owns a codebase: audits it, files and fixes issues, ships reviewed pull requests, keeps the tests green — gated by your merge rules.
Content & growth
Runs your content calendar — researches, drafts, and publishes on a schedule, in your voice — and keeps a backlog of ideas warm.
Bookkeeping
Keeps the books in a ledger it builds for itself: categorizes transactions, flags anomalies, and closes the month — every month.
Standing reports
Builds the report you need — pulling from your data and live sources — then runs it on a schedule and delivers the summary on its own. A weekly business review that writes itself.
A whole company
Give it a product idea and a budget. A CEO stands up engineering, growth, and support — each a self-running team. You're the board, not the operator.
…or anything else.
If the work flows through software you can connect (any MCP tool), a standing team can run it — and build its own tools when it needs them.
Not an assistant. A workforce.
An AI assistant does a task and waits for your next prompt. Jugg is a standing team you delegate a whole function to — and walk away from.
A single AI assistant
- Does the task you ask, then waits
- Works only while you're prompting it
- You bring it every piece of new work
- Forgets between conversations
- One assistant, one chat
- You drive every step
A Jugg workforce
- Runs a whole function you delegate
- Works around the clock, on its own
- Watches for new work and reacts itself
- Compounds memory in a shared Hive
- One team — scaling to a whole company
- You set the mission and check in
For a one-off task, a sharp assistant is often the better tool. Jugg is for the work you'd rather never have to think about again.
The always-on agent, grown up.
Running a self-hosted, always-on assistant — chat gateway, skills, memory, heartbeats, your own model key — is a category now. Jugg.ai is the next level of it: from one always-on assistant to an always-on organisation.
The always-on assistant you run for yourself is a solved problem. Jugg.ai is the always-on department you hand a job to — hosted, budgeted, approval-gated.
Limitless, not lawless.
The power of a swarm with the accountability of an org chart — memory, budgets, approvals, and a kill switch, built in from the first commit.
It runs without you
Not a chatbot you babysit. A standing team that keeps working — finishing one job, picking up the next — around the clock, while you sleep. And when it genuinely needs a call from you, it surfaces just that one decision instead of going quiet or guessing.
It reacts on its own
Put it on-call and it watches your inbox, tickets, or systems and acts the moment new work appears — no one has to ask. A support desk that answers email itself; an ops team that handles alerts as they land.
It never forgets
Every team keeps a Hive — findings, decisions, and facts with full provenance. It doesn't restart from zero; it compounds, getting better at your business the longer it runs.
From a team to a whole company
Give a bigger mission and a CEO fans it into departments — each a self-running team. Teams hand work to each other and stand up new teams as the work demands. Delegate one function, or an entire org that organizes itself. Same dial.
It builds its own tools
When your team keeps doing the same thing, it writes a tool for it — a real function, report, or integration — and asks you to approve the code before it runs. It can put those on a schedule, so a weekly summary or a watch-and-alert just happens. The longer it runs, the better-equipped it gets.
You hold the keys, always
Hard spend caps on your own provider key with an automatic halt, risky actions paused for approval, everything audit-logged, encrypted credentials — and a kill switch that stands the whole thing down in one click.
You bring the model. We run the organization.
Plug in your own API key — Claude, OpenAI, or a cheap high-volume model — and pay your provider directly at their price, with no markup from us. We charge a flat monthly fee for the part that’s actually ours: the orchestration, the memory, the skills, and the governance — priced by how many teams you run at once.
How pricing works
- 1
Bring your key
Add a model key from the provider you already use. It's stored encrypted and never shown to agents.
- 2
Pay your provider
Every token your teams use is billed by your provider, on your account, at their rate. We take nothing off the top.
- 3
One flat fee to us
A monthly subscription for the standing organization — durable memory, skills, self-built tools, approvals, recovery.
- 4
Hard caps you set
Every team has a lifetime spend cap and every job can have its own. At the cap it halts itself and asks you.
Plans
What a plan sets is how many teams run at once. Halted and archived teams don’t count, and every plan is plus your own model costs, paid to your provider.
Crew
NZ$14.99/month
2 running teams
One function, handed over. Two standing teams on call around the clock.
Studio
NZ$99/month
5 running teams
A few functions at once — support, ops, research — each with its own team and budget.
Group
NZ$399/month
20 running teams
A whole company: a CEO fanning missions into departments, twenty teams running.
You see every token your teams spend, itemised by team and by model. When models get cheaper, you keep the saving. Start with a 30-day free trial — no card needed, 2 running teams included. Prices are in New Zealand dollars and exclude any tax; the full amount is shown on the checkout page before you enter any card details.
Questions, answered.
What's it actually for?
Running functions, not doing one-off tasks. A support desk that watches your inbox and replies. An ops team that handles tickets as they land. A research desk that keeps a topic current. An engineering team that maintains a codebase. You set it up once; it runs and reacts on its own.
Do I have to babysit it?
No — that's the whole point. It runs continuously, reacts to new work without being asked, halts itself if something's wrong, and stays inside your budget. You open the dashboard when you want to check in, not because it needs you.
How is this different from just using an AI assistant?
An assistant helps you with a task, then waits for the next prompt. This is a standing team you delegate a whole function to: it persists, remembers, reacts to events, and scales from one team to a whole company — running operations hands-off, not answering questions.
Can it build things it doesn't already have?
Yes. When your team hits the same job repeatedly, it writes its own tool for it — a sandboxed function, report, or integration — and asks you to approve the code before it goes live. It can schedule those to run on their own, and a bigger team can even spin up new sub-teams and hand work between them. Its capability compounds: the longer it runs, the more it can do.
Is it safe to let it loose on my accounts?
It's never loose. It only reaches tools you explicitly connect, every action is logged, dangerous operations wait for your approval, and the kill switch halts everything instantly. Credentials are encrypted and never shown to agents.
Which AI model does it use?
Whichever you connect. Plug in a Claude, OpenAI, or low-cost provider key and the team runs on it. You can also mix: a strong model directing and planning, a cheap one doing the high-volume execution — which is how a big swarm stays affordable. Your key, your provider contract, your data residency.
What does it cost?
A flat monthly fee for the platform — Crew, Studio or Group, priced by how many teams you run at once — and you bring your own model key: Claude, OpenAI, or a cheaper high-volume model. Your model spend goes straight to your provider at their price; we don't mark it up or resell it. You set a hard budget your team cannot exceed and see exactly what it spends, so an idle on-call team costs almost nothing. When models get cheaper, the saving is yours.
How is this different from OpenClaw?
Same DNA, different unit. OpenClaw is an open-source, always-on assistant you run for yourself — it has triggers, memory, skills, your own model key and connections, and so does Jugg. The difference is what you're running: not one agent but a team with roles (a director that must delegate, managers, workers, departments, a CEO layer), governed by approval gates, hard budget ceilings per job, team and org, sandboxed self-built tools, per-tenant isolation and an audit trail — hosted, so there's nothing to run. If OpenClaw is the always-on assistant you run for yourself, Jugg.ai is the always-on department you hand a job to — hosted, budgeted, and approval-gated.
Stop running it yourself.
Let it run itself.
Your first workforce takes two minutes to hire. It never sleeps, never forgets, and never runs past the cap you set.
Hire your workforce