Ask anything about your plants — and let the assistant take the next step
The Mirox AI assistant lives inside the platform, right next to your data. It answers questions about production, events, tickets and access in plain language — and it can act: open a ticket, check a VPN connection, propose park data from a document. Every change to your data stays a proposal until you accept it.
Every question about your fleet, in plain language
The assistant runs inside Mirox, on the same live data your dashboards use. Ask “How are my parks doing?”, “Are there any problems?” or “Show my production over the last 7 days” and the answer is built from your own plants, portfolios and measurements — not from a generic knowledge base.
- Production, performance and revenue across plants and portfolios
- Events with their context and the estimated energy lost
- Data gaps, coverage and data volume per plant
- The Mirox documentation — and the documents you uploaded yourself
It can act — but never behind your back
The assistant is wired to the platform’s own functions, so a question can turn into work: it opens and comments on a ticket, checks whether a VPN connection is live, asks the plant’s on-site agent to run a fresh network scan, or invites a new member to your organization. It has no network path to your site itself — the scan runs on the agent there. And where your master data is concerned it never writes silently: it drafts a proposal you review, edit and accept.
- Tickets end to end — create, comment, link to an event, close or reopen
- Networking — VPN status, look up a device, a proposed fix for a broken connection
- Organization — members, roles, cooperations and “what am I allowed to do?”
- Park data — every change arrives as an editable proposal, never as a silent write
Where the assistant already helps
One conversation reaches across the platform — data, documents, tickets, network and access.
Production & performance
Ask for output over any period, per plant, portfolio or component — the same figures that back your dashboards.
Events & impact
Look up what happened, when, and how much energy it is estimated to have cost.
Documents & documentation
Search the Mirox documentation and read the files you uploaded to a plant — yield reports, datasheets, contracts.
Tickets
Create, update, comment, link, subscribe, close — the whole ticket lifecycle from the conversation.
Networking & VPN
Check VPN connections and their live status, look up a device, ask the on-site agent for a fresh scan, get router setup guidance.
Access & organization
List and invite members, change a role, invite a cooperation, and check your own permissions.
From a document to a reviewed park record
The AI Wizard turns your source documents into a list of proposed changes — reviewing them is the only interaction needed.
Pick the task
Onboard a new park from a yield report, verify a plant from the files already in its storage, upload a file and have it analyzed, tagged and sorted into the plant’s storage, or auto-tag a single file.
The AI reads the documents
It extracts the details and turns them into discrete proposals — basics, address, dates, business information, components, PR values and contacts.
You review and correct
Every proposed field is an editable input. Fix what is wrong before you decide; a proposal that depends on another waits for it.
Accept what you agree with
Accept or reject each proposal individually. Only what you accept is applied — and an apply that fails is shown as such, not hidden.
Beyond the chat window
Part of the assistant works without being asked — and its toolbox is open to the AI tools you already use.
Files tag themselves
A document uploaded to a plant is read, tagged and filed automatically — nobody has to start it, and the result stays visible and correctable.
Access sessions in plain language
Every remote-access session is condensed into a readable summary in the audit trail, so you do not have to reconstruct a log line by line.
Devices identified on discovery
A newly discovered device gets an identity suggestion from its fingerprint — a proposal for you to confirm, never a silent change.
Agentic access for your own tools
Connect your own external AI client under a scoped token: the same tool catalog as the built-in chat, inside the same permission boundary.
A glass box, not a black box
Every answer shows its work: which platform functions were called, with what input and what came back, plus the reasoning behind it if you want to see it. You choose which model answers, you can rate any answer — and the assistant says so itself: AI can make mistakes, so verify what matters.
- Expandable tool calls with input and result for every round
- Optional reasoning panel and a selectable reasoning depth
- Thumbs up/down and free-text feedback on any answer
- Attach images, PDFs, Office documents or text files to a question
Your model, your keys, your jurisdiction
By default the assistant runs on Mirox-operated infrastructure in Germany (EU), on a fixed model that is never trained or fine-tuned on customer data. If your organization would rather use its own AI accounts or a self-hosted endpoint, connect them and the assistant works through those instead — you decide which models your members may pick from, and everyone can see what their questions cost.
- Default: Mirox-operated infrastructure in Germany (EU), on a fixed model that never learns from your data
- Or bring your own — your organization’s AI accounts or a self-hosted endpoint, keys stored encrypted and never readable back
- Curate the model list your members choose from and set the organization default
- Usage and cost per member, plus an organization-wide overview for administrators
Why the demo above does not answer live
The chat above is the real assistant interface on demonstration data — the conversation view, attachments and transparency features exactly as in the product. The answers themselves come from the live data of your account and act with your permissions in your organization, so sending is reserved for signed-in users. What you see is exactly how it looks and works — and we will gladly walk you through it on your own plants.
Frequently asked questions
Who can use the assistant?
Every logged-in Mirox user finds it in the sidebar — it is part of the platform, not a separate product.
Where does my data go when I ask something?
By default to Mirox-operated infrastructure in Germany (EU), to a fixed model that is never trained or fine-tuned on customer data. An organization can instead connect its own AI accounts or a self-hosted endpoint; those keys are stored encrypted and can never be read back.
Can it change my park data?
Not on its own. It can do the things you can do in the platform, such as opening a ticket or asking the plant’s on-site agent to run a network scan — it has no network path to your site itself. Changes to park master data always arrive as a proposal you review, edit and accept before anything is written.
Can I see what it actually did?
Yes. Each answer lists the platform functions it called, with their input and result, and the reasoning behind it can be opened. You also choose which of the supported AI models answers, and you can rate every answer.
Why does the embedded assistant not answer my questions?
The demo shows the real interface, but the assistant answers from real account data and acts inside your organization — so live answers are reserved for signed-in users. In a walkthrough we show it working on your own plants.
Ready to Simplify Your Operations?
See how Mirox can bring your team together and save you money. Free pilot available — connect your first plant at no cost.