Sales
Pipeline pressure, next best action, and follow-ups that do not wait for someone to open a CRM tab.
United Kingdom · SMEs 20–250 people
Infrastructure you own and operate — not AI operations as a service. Intelligence, email, WhatsApp, calendar and apps, with Council and agentic modes. Licence free when you host; you pay for AI use. A services ladder when you need help.
What you get
aranduOS sits in the tools you already use — and across the commercial and operational work of a UK SME. One platform, not a fifth login for every department.
The personal layer: intelligence on the work, in the channels the leader already opens.
Intelligence
What is moving, what is stuck, what needs a decision — without chasing every thread.
A working inbox in the same flow, not another tab the assistant cannot see.
Messages, follow-ups, and a hand-off to you when it is sensitive.
Calendar
Time on the same map as the work — meetings the operation can read.
Apps
iOS, Android, Mac and Windows — desk, pocket and laptop.
The same Council and agentic modes across commercial and operational areas — with human gates where money or customers are at stake.
Pipeline pressure, next best action, and follow-ups that do not wait for someone to open a CRM tab.
Accounts, contacts and history in one context the assistant can use — not a silo beside the inbox.
Stages, stuck deals and who owes the next move — visible to the leader without a status meeting.
Drafts, revisions and chase across time zones — price changes wait for a person.
Buyer requests, account ops and overseas threads in the same flow as the rest of the desk.
Supplier follow-ups, orders and exceptions — what is late before it becomes a shortage.
Collections, approvals and exceptions that need a human — not autopilot on the bank.
Inbound cases, triage and drafts — escalate commercial exceptions instead of inventing promises.
Bookings, capacity and reminders for service businesses — calendar and WhatsApp in the same loop.
Agentic modes that understand, prioritise, act and ask for help — inside the limits you set.
How it thinks
Not a fixed if-this-then-that script. The system reads the situation, chooses the next step inside your rules, and asks a person when the stakes are high.
Council
The brain of the framework. It looks across the operation, weighs context, and recommends what should happen next — in business language, not a model dump.
Reads the case beyond a single rule: who is waiting, what changed, what still matters.
Orders the work inside the limits you set for the organisation.
Moves the routine in the systems and channels you already use.
Hands the sensitive step to a person. That is part of the intelligence, not a failure.
Money, a message to a customer, or anything irreversible waits for a human.
Trust and control
Operating AI on your business means knowing who can see what, which model may act, and what happened afterward. That is built into how aranduOS runs — not a slide after the sale.
Council and agentic modes run inside rules you set: what they may read, what they may draft, and what always waits for a person. Sensitive steps stay gated.
Access by security levels and profiles — roles decide which systems, cases and fields the assistant can use. Least privilege is the default, not an afterthought.
A trail of who asked, what the system saw, which model answered, and who approved the irreversible step. Ready for review, compliance and incident response.
Use cases
Built for UK SMEs of about 20 to 250 people — services and export to the world, with Council and agentic modes in the loop. Each case is a process, not a dashboard demo.
01
Consultancies and practices in the UK
Council ranks what is late. Agentic modes draft follow-ups. A message to the client waits for a person.
02
UK sellers shipping goods or services abroad
Inbox and WhatsApp stay in one flow. Follow-ups run agentically. Price or contract changes wait for a human.
03
Teams moving goods when something breaks
Intelligence shows what is stuck. The system does not invent a delivery promise without approval.
04
UK agencies delivering for overseas clients
Calendar and hand-offs sit with the job. Council recommends the next step in the pipeline.
05
Account and support teams with overseas buyers
One shared token pool for the firm — not a separate assistant login for every seat.
These workflows burn tokens on the Mirá multi-model gateway. Estimated monthly cost.
How you pay
We are not trying to win at “AI operations as a service.” We sell the infrastructure that lets a UK SME own and operate its AI business — with a services ladder when help is useful. That scales if the product stays easy enough to deploy and run.
UK companies with about 20 to 250 employees — big enough for real operations, small enough that another managed AI team is the wrong default.
Someone else builds and runs the agents for you. Useful for some firms — and a different product. You rent the outcome; you do not own the stack.
You host or we host. You run the workflows. You pay for AI tokens. Setup, integration and support are a ladder you climb only when you choose — not the only way in.
A 20–250 person UK firm hears “AI platform” and expects another software subscription. Instead: licence free, host on your infrastructure, pay only for AI tokens, optional configuration and support. That lowers the door.
Put aranduOS on the server you choose. Hosting cost is yours. No Mirá maintenance fee for the box. You pay us for the tokens the install uses.
Less operational worry, higher cost. We run the stack. Extra services are still not included in the base — you add them if you need them.
Required from Mirá
Services ladder (optional)
Documentation is free. Hands-on support and the extras above are contracted. Local help on the ground sits with the UK ecosystem. The aranduOS name stays on — no white-label.
Token costs
aranduOS routes work across several models. You are billed in BRL on the Mirá wallet. Figures below are indicative GBP for a team of ten sharing one pool.
| Model | Est. £/mo typical | R$/mo | In / out per 1M | Light–heavy £ |
|---|---|---|---|---|
| gpt-oss-20b | £2.0 | R$15 | R$0.24 / R$0.90 | £0.66–£5.3 |
| Qwen3.5-9B | £2.8 | R$22 | R$0.60 / R$0.90 | £0.95–£7.6 |
| Mistral-Small-3.2-24B | £3.9 | R$30 | R$0.54 / R$1.68 | £1.3–£10 |
| gpt-oss-120b | £4.9 | R$37 | R$0.48 / R$2.40 | £1.6–£13 |
| gpt-4o-mini | £7.8 | R$59 | R$0.90 / R$3.60 | £2.6–£21 |
| Meta-Llama-3.3-70B | £16 | R$121 | R$4.02 / R$4.02 | £5.3–£42 |
| gpt-4.1-mini | £21 | R$158 | R$2.40 / R$9.60 | £7.0–£56 |
| Qwen3.6-27B | £31 | R$238 | R$2.40 / R$16.20 | £10–£83 |
| Qwen3.5-397B-A17B | £43 | R$324 | R$3.60 / R$21.60 | £14–£114 |
| o3-mini | £57 | R$436 | R$6.60 / R$26.40 | £19–£153 |
| o1-mini | £57 | R$436 | R$6.60 / R$26.40 | £19–£153 |
| gpt-4.1 | £104 | R$792 | R$12.00 / R$48.00 | £35–£278 |
| gpt-4o | £130 | R$990 | R$15.00 / R$60.00 | £43–£348 |
Buying one Claude Team or ChatGPT Business Standard seat per person for ten staff is about $250/month at list. aranduOS keeps one licence-free install and a shared token pool. The comparison uses a mid mix — mostly mini and o-series, gpt-4o only where it earns its keep — at least 50% below an all-gpt-4o bill.
Mixed routing (45% gpt-4.1-mini · 35% o3-mini · 20% gpt-4o)
Light–heavy £19–£148 · about 57% below all-gpt-4o
Standard seat, billed monthly (list price). Not a quote from those vendors.
Charged in BRL. GBP uses indicative FX (USD/BRL 6, GBP/USD 0.79). Gateway list rates as of 6 October 2026; markup today 0%. Usage varies.
Product layers
The personal layer.
Most teams start here
The core pitch.
Where data sits
Self-hosted is not open source and not air-gapped. Local screens and operational data stay on the install you control. AI goes through Mirá when you use it.
One installation per end customer. You or a certified integrator operate the box. Licence free; hosting bill yours; tokens to Mirá.
We run the platform. You use the product. Extras (setup, consulting, support, email) are still optional.
Who does what
Mirá engineers the product, the releases, this site and the LLM gateway. Organisations in the UK ecosystem help companies implant and support locally — without a Mirá maintenance fee on Self-hosted.
Mirá
Product, engineering, releases, optional hosted stack, LLM gateway.
UK ecosystem
Local support and implantation in the United Kingdom.
You
Host on your servers (licence free + tokens), or ask us to host — and which optional extras you want.
Before you write
Council is the brain of the framework: it analyses the operation and recommends the next step. Agentic modes understand, prioritise, act and ask for help — inside limits you set. Sensitive actions wait for a person.
No. We are not selling AI operations as a service. aranduOS is infrastructure you own and operate — free licence when you host, pay for AI tokens, climb a services ladder only if you want help.
UK SMEs of about 20 to 250 employees — services and export firms that want to run AI operations themselves, not rent another managed team.
Rules for what Council and agentic modes may do; managed access by security levels and profiles; a complete audit trail of asks, model use and human approvals.
No. On Self-hosted the licence is free. You pay for your own hosting. The one mandatory purchase from Mirá is LLM tokens through our gateway.
Mandatory: LLM usage (tokens) via the Mirá gateway. Optional: setup, configuration, consulting, support, email accounts, and other extras. Hosted by Mirá is the other way to run it — less worry, higher cost; extras still not in the base.
No. We can host it for you. That costs more and removes the server work from your side. Either way, extras are contracted separately.
Neither. Closed-source images you run yourself. Local operation can work without calling us. AI uses the Mirá gateway when enabled — that is how tokens are billed.
Operational data in that install stays on the infrastructure you choose; you are the controller. LLM (and optional mail) traffic goes through Mirá under the terms of that use. This marketing site is separate — see the privacy page.
Mirá builds and maintains aranduOS. Local support comes from the UK ecosystem. Hands-on support from Mirá is an optional contract; documentation is free.
iOS, Android, Mac and Windows, plus WhatsApp, email and calendar in the same flow.
NEXT STEP
Headcount, revenue, sector, company type, Companies House number, fixed telephony and paid AI seats. We use that to estimate savings and the next step.
Company size
People, annual revenue, sector and company / tax type.
Telephony and AI
Inbound numbers, extensions, provider and paid AI licences.
Savings view
We come back with an estimate and how to implant.
Prefer email? [email protected]