AI Center is the AI layer of your Ubidots account: a chat that answers questions from your real device data, plus AI agents that you configure, connect to other systems and share with your end users on the web or on WhatsApp. It is included in the Professional, Industrial and Enterprise plans, and usage is measured in AI credits.
Think of an agent as an analyst that already has your data: it reads your devices, variables, events and incidents directly from Ubidots, so you ask a question in plain language instead of building a dashboard or digging through device lists.
What's in AI Center
Chat. Ask questions about your operation in plain language and get answers based on the data in your account. See Use Chat below.
Agents. AI assistants you configure: you choose the model, write the instructions (system prompt) and decide which connectors each one can use. See Create and Configure AI Agents.
Watchers. Run an agent on a schedule: a Watcher sends a prompt to an agent at set times, so the agent checks your operation and reports without anyone having to ask.
Connectors. Give agents data and tools from other systems, such as weather services, solar inverter platforms or any MCP server. See Connectors.
Channels. Decide where your end users talk to your agents: the web console inside your Ubidots app, or WhatsApp. See Channels and the Web Console and WhatsApp Channel.
Every agent can use your Ubidots data from the start: it looks up devices, variables, last values, statistics, events and incidents through the built-in Ubidots connector. It always works within the permissions of the user it is answering, so it only sees what that user can see.
Using your own AI tools? To query Ubidots from Claude or another AI client outside Ubidots, use the Ubidots MCP Server instead. It doesn't use AI Center or AI credits. |
Requirements
Plan: Professional (including the Professional trial), Industrial, Scale or Enterprise.
Role: account owners and admins open AI Center and configure agents, connectors and channels. End users don't see AI Center's configuration; they chat with the agents you give them through a channel.
Use Chat
Open Chat
From any page in your Ubidots account, click AI Center → Chat in the navigation menu. The welcome screen, Welcome to AIoT Chat, opens with the built-in IoT Fleet Assistant selected and suggests questions under Try asking:. Click one to run it without typing.
Ask a question
Choose an agent. Click the agent's name below the message box to open the agent menu, and pick one. Agents you created appear under My Agents; the ones Ubidots provides, such as IoT Fleet Assistant, appear under System Agents.
Start a conversation. Click New chat. Your previous conversations stay in the sidebar, grouped by date, so you can go back to them.
Ask in plain language. Type your question and click the arrow button to send it. With IoT Fleet Assistant, the Add Device button below the message box lets you attach specific devices to the question.
Follow up. Keep asking in the same conversation to drill into a device, variable or time range the agent mentioned.
Rate the answer. Use the thumbs up or thumbs down under an answer to tell us whether it helped. Copy message copies the answer.
Tip: Name the device, variable and time range when you can. A specific question is answered faster and uses fewer AI credits than one that makes the agent scan your whole fleet. |
Each conversation is independent: context from one conversation isn't carried over to a new one. To remove a conversation, use Delete conversation in the sidebar.
What you can ask
"Which devices haven't sent data in the last 24 hours?"
"How did the temperature in cold room 3 behave this week? Any readings out of range?"
"Compare the average energy consumption of the three plants last month."
"Summarize the open incidents and who they're assigned to."
"When did pump-12 last report, and what were its last values?"
Plans and limits
Each plan includes a number of AI agents and a monthly allowance of AI credits. The hard limit is the most AI credits the account can use in one billing period.
Plan | AI agents included | Maximum AI agents | AI credits included per month | AI credits hard limit per month |
Professional | 1 | 3 | 1,000 | 10,000 |
Professional trial | 1 | 1 | 1,000 | 1,000 |
Industrial | 3 | 5 | 10,000 | 100,000 |
Enterprise | 5 | 10 | 50,000 | 100,000 |
Additional AI credits
Usage above your included credits is billed at $10 USD per 1,000 AI credits, in blocks of 1,000, up to your plan's hard limit. On the Professional trial the hard limit equals the included credits, so there are no additional credits.
Additional agents
You can create agents beyond the ones your plan includes, up to the plan's maximum. Each additional agent costs $29 USD per month.
Hard limit and warning emails
We send a warning email when usage reaches 90% of the hard limit and again at 100%. When the hard limit is reached, AI requests stop until your next billing period starts. If you need a higher limit, contact us.
How AI credits work
Every message an agent answers uses AI credits, whether it comes from Chat, the web console or WhatsApp. The amount depends on how many tokens the message processes and on the model's multiplier:
AI credits = tokens processed × model multiplier ÷ 1,000, rounded up. Every message uses at least 1 credit.
What counts as tokens
Tokens are the units an AI model reads and writes, roughly three quarters of a word each. For every message, the agent processes:
Your question and the agent's answer.
The agent's system prompt and the conversation so far. The agent re-reads earlier messages on each turn to stay in context, so the 10th message in a long conversation costs more than the first.
The data the agent looks up to answer, such as device lists, values and statistics. A question about one device fetches less than one about your whole fleet.
An agent can look up data several times before it answers, up to six model calls per message. The tokens of every call add up.
The model multiplier
Each model has a credit multiplier, shown in the agent's model picker as Credits: N× per token. Faster, lighter models cost less per token; the most capable models cost more. Multipliers are 0.25×, 0.5×, 1× and 2×.
For example, the same message that processes 6,000 tokens costs:
Model multiplier | AI credits |
0.25× | 2 (1.5, rounded up) |
0.5× | 3 |
1× | 6 |
2× | 12 |
Tips to use fewer credits
Be specific. Name the device, variable and time range so the agent fetches only what it needs.
Start a new chat when you change topic. A fresh conversation doesn't re-read the old one.
Match the model to the job. A model in the Fast group handles routine checks well; keep Premium models for analysis that needs them.
Keep system prompts focused. The system prompt is sent with every message.
Attach only the connectors an agent needs. Each connector's tools are described to the model on every request.
Note: AI credits are a single monthly pool for the whole account. Chat, your agents on every channel and the AI Coding Assistant all draw from it. |
Privacy
Agents use zero data retention (ZDR) by default: prompts, device data and responses are processed to answer each request and then discarded, and the AI model provider keeps no copy. Whether ZDR is on or off, your data is never used to train AI models. See Create and Configure AI Agents for details.
Related articles
AI Center: Connectors
AI Center: Channels and the Web Console
AI Center: WhatsApp Channel