A new site
One place that shows the platform as it is today
Thinger.io has a new website. It shows the platform as it works today: the devices it connects, the operation it keeps in view and the AI agents that now work on it with you.
The platform has grown in many directions: a Marketplace of device integrations, remote access to Linux devices in the field with ThinRemote, its own device protocol published as an open Internet-Draft, and an MCP server for AI agents. The new site puts all of it in one place, with one language, so you can see in a few minutes what the platform does and where your project fits.
This post walks through what changed for visitors, and then spends most of its time on the part that changed the most: how AI agents fit into the platform, and what that means for a team that runs devices.
What you will find
What changed for visitors
Clearer menus, a home that shows an operation at work, and every edition on one path.
A home that shows an operation
Menus by what you look for
Every edition on one path
A menu made for phones
A page for each part
The blog, with its history
A home that reads like the life of a project
The home is laid out as a walk through a project. Near the top it shows a dashboard in use, then the agent that built it, and from there it follows the order in which a project grows: start with one device, scale, plug in hardware you already have, extend the platform and carry it in your pocket. Each part has its own short label:
The home, top to bottom
- Plug in
- Skip the integration work
- Extend
- More than a console
The case studies, the newsletter and a way to reach the team close the page.
The editions read as one path. A free account on the community cloud is where most projects begin; Private Cloud runs the full platform on a dedicated instance in the AWS region you choose, managed by us, on the plans of the pricing page; On-Premise and Enterprise cover the projects that must stay inside your perimeter or under a custom contract; and White Label / OEM lets integrators and device makers sell it under their own name. Your devices, dashboards and integrations work the same in every mode, so moving up never means starting over.
The blog you are reading is part of the change too. Every post, back to the first ones of 2016, is now a page of its own, sorted into Product, Engineering, Insights, Events, Recognition, and Projects & grants. The archive stays as it was written: older posts keep their date and carry an update note where the platform has moved on, so you know what still holds.
The focus
Why AI sits at the center of the platform
An agent is only as good as its hands. A model can reason about pumps, but it cannot see yours.
Most of the new site turns around one idea: an AI agent connected to the platform can see your fleet, act on it within the limits you set and build what your team needs to look at. You ask in plain language and get answers from live data. The agent proposes the fix, writes the rule or builds the view, and only acts with the permissions you give it.
Two pieces make that work, and each has a page of its own on the site: custom dashboards that an agent writes from a description, and an MCP server that lets the agent read and operate your account. You can see both on the home: the water network dashboard at the top is the one the agent console below it was asked to build.
Custom dashboards
Describe the view, and an agent builds it
A view like the one on the home used to take days of frontend work. Describe it to an agent and it is yours in minutes.
Describe the view you need and an AI agent writes it as a complete web app that runs inside your console, on your live data. It is not a layout of widgets: it can have hierarchies, navigation, several pages and a design of your own. A map of the network, a plant schematic or a portal for your customers: whatever your team needs to see, laid out their way.
From a request to an app in your console
An AI agent writes the app
Into a file storage
Opened in the console
Your session, no second login
Widgets on a grid
- No code and no agent: drop widgets on a grid and point them at your devices
- Time series, gauges, maps, tables, values, controls and custom HTML
- Shared through a public link governed by an access token
A dashboard in minutes
An app an agent wrote
- Any technology that runs as a web app in the browser
- One app as a standalone view or as a template per device or per product, with drill-downs into the console
- To change it, ask the agent again: another view, a new chart, a different layout
Minutes, not a sprint
Both run on the same live data, and both stay: widgets for something quick, an agent for the view built around your operation. See Dashboards, where one solar site is shown both ways.
AI & MCP
An agent that operates your fleet with your permissions
The Thinger.io MCP server, run by the Thinger CLI, gives your AI agent the documented API: ask in plain language, get answers from live telemetry, delegate routine operations.
The Model Context Protocol is the open link between the agent and the platform. Claude, Codex, Cursor or any other MCP client connects through the Thinger CLI, with a token you choose, and there is nothing to install on the devices: the platform keeps talking to them as it always has. What the agent gets is a set of typed tools, in three groups.
Read: see the fleet
thinger_resource_list, thinger_bucket_data_read, thinger_device_overview. Act: operate devices
thinger_device_resource_call, thinger_resource_update, thinger_resource_call. Build: shape the platform
thinger_resource_create, thinger_dashboard_publish. Here is what that looks like in a short session, following a station that went silent in the night: the agent reads the history, acts on a device only after the operator confirms, and builds what was missing.
you PS-07 went silent at 02:14. What happened before it dropped? bucket_data_read · resource_list agent Battery fell from 3.9 V to 3.1 V through the evening while flow stayed normal, and no alarm fired. It points at the power supply, not the pump. you Can PS-08 cover its zone until someone gets there? device_overview · device_resource_list agent Backup pump B at PS-08 is idle and healthy. Start it now? you Yes, start it. device_resource_call agent Pump B is running at 12.4 l/s. you Make sure we hear about this earlier, and give the night shift a view of it. resource_create · dashboard_publish agent Drafted a rule for any station battery under 3.4 V, left disabled for your review. The Station power dashboard is up, with seven days of battery per station.
Illustrative, from /platform/ai: written for the page, not recorded
The agent gets a key, not your keys
Agents authenticate like any third party application: with an access token you create, scope and revoke. A token can allow reading one bucket and nothing else, or operating a single device's resources; reading and acting are separate grants. The agent's key is not your login, and it can be disabled or deleted from the console whenever you choose. There is no separate agent backdoor to audit: every call an agent makes is an ordinary authenticated API request, subject to the same permissions as any other client.
A sober starting policy, as the AI & MCP page puts it:
Read only first
Actuation, resource by resource
A human confirms
For your team
What this means for a team
Less time getting to the data, more time deciding what to do with it.
Operators ask instead of digging
"Which pumping stations had pressure spikes last night?" is a question, not a report to build. The agent reads the devices, the stored data and the alarms, and answers with the station that matters and the reading behind it, then proposes the next step for you to approve.
The view fits the job
The night shift, a plant manager and a customer each need a different screen. Each one can have its own, built on the same live data, and when the need changes, the agent that built it changes it.
Routine work is delegated, not handed over
Start read only, allow actions later, revoke any time. A rule can be drafted and left disabled for review, a rollout can start with a dry run, and a person confirms anything that moves water, power or money.
The same devices, one more way in
A pin you declare as a resource on the board can be switched from a dashboard, the REST API or an AI agent. The agent is one more client of the platform, not a separate system to run beside it.
The agent proposes the fix, writes the rule or builds the view, and only acts with the permissions you give it.
Underneath
The same platform underneath
The agents work on the platform that was already there, and the site now shows that too.
Devices connect over IOTMP, the protocol that has carried Thinger.io's device traffic since 2015, or over MQTT, HTTP, LoRaWAN and Sigfox. The IOTMP page is redesigned around one idea: telemetry, remote calls, streams, terminals and tunnels share one TLS connection, and the protocol is now an open IETF Internet-Draft that anyone can implement.
The Marketplace holds 170+ device integrations from 40+ manufacturers and 4 network servers: find your model, install it from the console, and readings arrive decoded, with storage and a starter dashboard already set up. And the mobile app opens the same devices, projects, dashboards and data streams as the web console, free on iOS and Android.
Where to start
Start with one device and one question
Create a free account, connect a device and point your MCP client at it with a scoped token.
The free Cloud account is free forever, with fair limits, and no credit card is required. Ask the agent something about your own data, then ask it for the view you would want on a screen. When the project needs its own instance, your servers or your brand, the editions are on the pricing page, and the team is one message away.
Put an agent to work on your fleet
Start on the free cloud, or tell us about the operation you want to run with an agent beside you.
