English 
Français 
Español 
Deutsch 
Português 
Italiano 
العربية AI Assistant
The AI IoT platform where the assistant does the work
Most "AI" in software is a chat box bolted onto a help page. Kilo's assistant works like an experienced IoT integrator by your side — it knows your whole deployment, answers from your live telemetry, provisions devices, writes and tests automation rules, and configures alarms with escalation. It does the work, not just the talking — and before anything consequential happens, it asks you first.
Included in every plan · Free for 5 devices
app.kiloiot.io · AI Assistant
Real productGrounded answers
An AIoT platform assistant grounded in your live data
Every answer about your devices, rules, alarms, and dashboards comes from your real deployment — read through the platform at the moment you ask, scoped to your permissions. If something can't be retrieved, the assistant says so instead of inventing an answer.
There is no query language to learn. Ask the way you would ask a colleague — "which freezers breached −18 °C last night?" — with natural time expressions like "yesterday" or "last 3 hours", and refine with follow-ups without restating anything.
Live status
"Which devices are offline?" — current readings, battery, signal, last seen.
Historical analysis
Aggregations, threshold violations, and period-over-period comparisons across your full retention window.
Inline charts
Ask for a line, bar, pie, or scatter chart and it renders right in the chat.
Conversation memory
Each follow-up builds on the previous context, so complex analysis stays conversational.
It acts, not just advises
From question to IoT automation: it builds, tests, and deploys
Describe the behavior you want and the assistant authors the complete automation rule — including the CEL expressions — then simulates it against a matching and a non-matching value to prove it fires correctly before deploying. You see every step, and you can refine it conversationally: "make it fifteen minutes", "also notify the facility manager".
Anything destructive or hard to reverse is gated behind an explicit Confirm Action / Cancel prompt, and after every change the assistant re-reads the result to verify its own work.
app.kiloiot.io · AI Assistant
Real product
Provision devices
Give it a DevEUI and AppKey and it onboards the LoRaWAN device, then runs diagnostics to confirm data is arriving.
Write & test rules
Authors the full rule with its CEL logic, simulates both outcomes, and deploys it.
Create alarms
Sets up alarm definitions with severity, recipients, channels, and multi-step escalation chains.
Recommend hardware
Suggests compatible sensors from the partner catalog — real, fitting options, not a generic list.
Open by design
AI assistant pricing: a monthly allowance — or bring your own model
The assistant is part of the platform, with a monthly request allowance that scales with your plan. Your remaining allowance is always visible above the chat input, and when you reach it, the rest of the platform is unaffected.
Prefer no limits? Connect your own model key — OpenAI, Anthropic, Ollama, or any OpenAI-compatible endpoint — and your model serves your chats with no allowance at all.
Under the hood, the assistant runs on an enterprise-grade agent runtime built in-house — and open-sourced as Synthetic Brew, so you can see exactly how it is engineered.
github.com/syntheticinc/syntheticbrewPrivacy & security
IoT data privacy: it sees what you see — nothing more
The assistant authenticates with your active session and inherits your exact permissions and organization context. It cannot see more data than you can, and it cannot perform any action your role does not allow.
Permission inheritance
Read-only users can query freely, but the assistant cannot change anything on their behalf.
Organization isolation
Strictly scoped to your current organization — it never crosses tenant boundaries.
Confirmation gates
Deleting a device or rule, or resolving an alarm, always pauses for your explicit approval.
Telemetry stays home
Only your question text goes to the language model — device telemetry is not sent to external services.
Novice or expert
An experienced IoT integrator by your side
Onboarding fifty sensors used to mean fifty registration flows. Now you describe what you're adding and the assistant does the mechanical work — provisioning, rule drafts with tested CEL, alarm chains, and hardware shortlists — while you keep approval over every consequential change. It helps whether you're new to IoT or a seasoned integrator: it walks you through setup, explains features, troubleshoots, and searches the platform knowledge base and IoT references. Experts keep CEL programming for full customization; beginners let the assistant do the heavy lifting — no code required.
Book a callFAQ
AI assistant FAQ
Is this just a chatbot on top of documentation?
No. The assistant reads your live deployment — device state, telemetry history, rules, alarms — at the moment you ask, and it can act: provision devices, author and deploy rules with their CEL, and configure alarms with escalation. Documentation search is just one of its sources.
Can it change things without my approval?
Consequential or destructive actions — deleting a device or rule, resolving an alarm — always pause for an explicit Confirm Action / Cancel prompt. It also only ever acts within your role's permissions: what you cannot do, it cannot do.
What happens when I reach my monthly AI allowance?
The rest of the platform is unaffected. To keep using the assistant, upgrade your plan or connect your own model key (OpenAI, Anthropic, Ollama, or any OpenAI-compatible provider) — your own key is not subject to the allowance.
Is my device data shared with model providers?
Only the question text is sent to the language model for interpretation. Your device telemetry stays within the platform and is not transmitted to external services for processing.
Ask your deployment a question
Start free with 5 devices, open AI Chat, and ask what your sensors are doing — no payment details required.