> For the complete documentation index, see [llms.txt](https://docs.chirpwireless.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.chirpwireless.io/changelog.md).
# Changelog
Chirp changelog — Flight Log entries for every release, with feature highlights and links to the matching guides.
Flight Log. Release 3.9.0
Chirp automations send alerts, check other readings, or control equipment when something happens at home. A **trigger** is a saved condition that starts an automation. It can now watch one device or several selected devices, and it can react immediately or only after the condition lasts for a set time. Before 3.9.0, covering nine windows meant maintaining nine automations; now one trigger and automation can cover them while keeping each window's condition and countdown separate. This release also puts metrics on one searchable page, prints exact values on bar charts, improves the Helper and bring-your-own-AI connection, and helps connected AI apps tell read-only actions from actions that change something. [app.chirpwireless.io](https://app.chirpwireless.io)
***
#### What's in This Release
* **One automation for up to 500 devices** — Previously, applying the same automation to nine devices required nine separate automations. Now one trigger can select those devices, check each one separately, and start the shared automation for whichever device meets the condition.
* **Wait before an automation runs** — A trigger can require a condition to remain true for 10 seconds to 30 days before starting the automation. For example, it can ignore a refrigerator door opened briefly but act when the door remains open for 10 minutes.
* **All sensor metrics on one page** — A metric is a type of reading, such as temperature, humidity, or battery level. Metrics that were spread across three tabs are now available in one searchable list at **Devices → Metrics**, where you can also add and edit them.
* **Show values on bar charts** — **Display value on bar** prints each value on its bar. **Show metrics below** adds the current readings beneath the chart.
* **Use your own AI account** — Connect an OpenAI, Anthropic, OpenRouter, Ollama, or compatible account instead of using the monthly Helper-message allowance included with your Chirp plan. Ollama model names with version tags are now supported.
* **Connected AI apps can distinguish read and write actions** — Actions available to ChatGPT, Claude, and other connected AI apps now identify whether they only read information or can change or delete something.
* **Reliability and interface improvements** — Sensor photos save correctly, the Activity Log refreshes after permission changes, widget colors work again, log retention follows your plan, and the Helper reports completed actions more reliably.
***
**One automation for many devices**
An automation is a flowchart that tells Chirp how to respond to sensor data. It can check readings, send an alert, or control equipment. A trigger defines the condition that starts the automation.
Before 3.9.0, each automation could listen to only one device's sensor. If the same condition and response applied to nine windows, you needed nine separate automations. Updating the condition later meant editing every copy.
Now one trigger can select up to **500 devices**, and one automation responds for all of them. Adding another sensor means updating that trigger instead of copying the automation. This selection is saved inside the trigger rather than as a reusable device group.
Most multi-device triggers are simple: every selected device supplies the same reading and is watched independently. Every watched device has its own trigger state and countdown, so one window remaining open does not affect any other window.
A selected device can also supply a shared reading instead of being watched. For example, nine window devices can each supply their own open-or-closed reading while one thermostat supplies the heating status for every window check. Before you save, **How this trigger will run** shows one row for each watched device and identifies any shared reading in the **Uses** column.
An alert can also identify the watched device that triggered the automation. Include `vars.device_name` in the alert message so it says “Landing window is open” instead of only “A window is open.”
[→ One Automation for Multiple Devices](/rules-engine/going-deeper/triggers/multiple-devices.md)
***
**Make an automation wait before it runs**
Before 3.9.0, an automation ran as soon as a sensor reading crossed the limit you set. An immediate response is useful for urgent conditions such as a water leak, but a brief increase in humidity or an open door does not always require action.
A trigger can now require the condition to remain true before the automation starts. Choose **Only if it lasts**, then set a duration in seconds, minutes, hours, or days. New triggers start at 10 minutes; you can set any duration from 10 seconds to 30 days.
For example:
* Ignore a refrigerator door opened briefly, but send an alert if it remains open for 10 minutes.
* Ignore the temporary humidity from a shower, but act if the bathroom remains at 80% humidity all day.
* Ignore a few seconds of movement near a gate, but respond if motion continues for 10 minutes.
Two additional behaviors control how the timer works:
* **A gap between sensor reports does not reset the countdown.** A battery-powered sensor that reports every 15 minutes will not cancel a 20-minute wait simply because no new reading arrived between reports.
* **Clearing the trigger can have its own condition and delay.** Under **Clear behavior**, define when the trigger returns to normal and how long that clear condition must last. This prevents one normal reading from clearing a condition too early.
[→ Trigger Timing](/rules-engine/going-deeper/triggers/trigger-timing.md)
***
**All sensor metrics on one page**
A metric defines a type of sensor reading, such as temperature, humidity, battery level, or whether a door is open. The definition tells Chirp the metric's name, unit, and data type so readings from different manufacturers can be used consistently in dashboards and automations.
Before 3.9.0, metric definitions were divided across three tabs. They are now managed in one list at **Devices → Metrics**. From this page, you can search by name, filter by unit or type, sort by newest or oldest, and add or edit a metric in the same dialog.
Metrics apply across your Chirp account. Editing one changes it for every sensor that uses it, so Chirp now warns you about the wider effect and asks for confirmation before saving.
[→ Data Templates](/devices/data-templates.md)
***
**Show values on bar charts**
A chart widget displays sensor readings on a dashboard. Before 3.9.0, values on a bar chart had to be estimated from the chart's scale.
Turn on **Display value on bar** to print the value directly on each bar. This option is available only for bar charts.
Turn on **Show metrics below** to add a row beneath the chart with each metric and its current reading. This can make compact dashboard widgets easier to read.
[→ Bar Chart](/dashboards/adding-widgets/chart-widget/bar-chart.md)
***
**Use your own AI account**
Each Chirp plan includes a monthly allowance of Helper messages. You can instead connect an account from OpenAI, Anthropic, OpenRouter, Ollama, or a custom OpenAI-compatible provider. Messages sent through that connection do not use the allowance included with your Chirp plan.
Open **Connect your AI** at the top of AI Chat and enter the provider, endpoint, API key, and model name.
Before 3.9.0, Chirp rejected Ollama model names that included a version after a colon, such as `gemma4:31b`. The suggested Ollama models were also unavailable to free accounts. Versioned model names are now accepted, and the suggestions have been updated to models that work with a free Ollama account.
Connection errors now explain whether the provider rejected the API key, the account does not include access to the selected model, or the model name does not exist. After you connect, the panel continues to show which model is in use.
[→ Your Chats and Your Own AI](/ai-assistant/chats-and-your-own-ai.md)
***
**Connected AI apps can distinguish read and write actions**
Chirp can connect to AI apps such as ChatGPT and Claude. These apps can answer questions about your sensors and, with permission, perform actions such as controlling equipment.
Before 3.9.0, the available actions did not have clear titles or identify whether they only read information or changed something. The AI app could not reliably distinguish a question such as “How warm is the nursery?” from an action such as “Turn off the heating.”
Each action now identifies whether it reads information, changes or deletes something, or reaches outside your Chirp account. Compatible AI apps can use this information to answer read-only questions directly and request confirmation before performing a change. No additional Chirp configuration is required, and the published action list is generated from the same definitions used by the server.
[→ MCP Server](/api/mcp-server.md)
***
**Reliability and interface improvements**
The Helper now reports the outcome of an action more accurately:
* It confirms a reporting-interval change only when the sensor accepts it.
* Retrying sensor registration no longer replaces the name you chose.
* Sensor registration no longer adds an unrequested temperature mapping.
* The Helper correctly explains that Chirp can control equipment. When an action must be completed in the web interface, it directs you to the correct screen.
* A successful sensor registration no longer displays a contradictory failure message.
* Confirmation messages display line breaks correctly.
* Replies no longer repeat the same answer in both text and a widget.
* Replies that arrive all at once appear immediately without requiring a page refresh.
* Sensor limits are enforced consistently whether a sensor is added through the Helper or through a form.
Additional improvements in this release include:
* Sensor photos save correctly.
* The Activity Log refreshes after permission changes, so new entries appear without reloading the page.
* Last Data widget values change color according to their conditions again.
* The Key Vault describes only the fields Chirp uses.
* Device-log retention follows the limit included with your plan.
* The Helper can retrieve longer periods of sensor history.
* Sensors that report once a day are no longer marked offline after only a few hours, and the warning identifies its units.
* The support form lets you select a bug report, feature request, or integration request.
* Emulated sensors timestamp their first reading correctly and display properly on mobile screens.
* The trigger dialog fits on mobile screens.
* Empty states are consistent across the platform, including automation artifacts. Device and gateway empty states also link to the shop when no hardware has been added.
* The Download the app card displays correctly.
Flight Log. Release 3.8.0
Three big ones in 3.8.0. **First, you don't have to wait for the parcel** — **pretend sensors** let you build your dashboards and your alerts today, test that they really go off, and swap in the real sensor when it lands. **Second, the bit people have asked for since the helper arrived**: say *"turn on the lamp"* and it **turns on the lamp**. Your helper works the house now, it doesn't just watch it. **And third — the one we're most excited about — connect ChatGPT or Claude and they can switch things on too.** The AI app you already use, controlling your actual home, asking you first every time. There's also a **What's New** note in the app so you stop missing things like this. [app.chirpwireless.io](https://app.chirpwireless.io)
***
#### What's in This Release
* **Pretend sensors** — Set up your home before the hardware arrives. Make up readings, watch your alerts fire, then point the same device at the real sensor when it turns up and keep everything.
* **The helper can switch things on** — *"Turn on the lamp"* now turns on the lamp. Ask it to flip a relay or change how often a sensor reports, and it does — showing you what it's about to do, asking first, then telling you whether the device got it.
* **Control your home from ChatGPT or Claude** — Connect the AI app you already use and it can switch things in your house too, with the same confirmation before anything happens.
* **The helper sets up pretend sensors as well** — Ask for one and it builds it, sends a test reading, and takes it live onto your LoRaWAN connection later.
* **What's New in the app** — A note that tells you what changed, so a release doesn't quietly pass you by.
* **Smoothing things out** — Your plan dates read correctly, adding a widget stops shoving your other widgets around, long addresses don't run off the screen on your phone, and a handful of labels finally say what they mean.
***
**Set your home up before the sensors get here**
Ordering a sensor used to mean a few days of nothing. You could not lay out a dashboard because there was nothing to put on it, and you certainly could not check your alert worked.
Now you can add a **pretend sensor**: pick the model you have ordered and it will make up believable readings all by itself, on whatever schedule you choose. Your dashboard fills in, your automations run, and your alerts go off — all before the box arrives.
The best part is testing an alert honestly. You want to know the damp alert really lands on your phone at 2am? You do not need to leave a tap running. Park the humidity at 85%, or fire a single reading, and watch what happens.
Then Thursday comes, the sensor arrives, and you open the same device and switch it from pretend to real. The dashboard, the alerts, who gets told — all still there. It takes five minutes instead of an evening.
[→ Pretend sensors](/connectors/emulator-connector.md)
***
**"Turn on the lamp" — and it turns on the lamp**
Your helper could already answer anything about your home and set up automations for you. What it could not do was actually *do* anything, which is a strange thing to explain to someone standing in their kitchen.
That changes here, and it's the biggest thing we've shipped for the helper yet. Ask what a device can do and it tells you; ask it to do one of those things and it does it. Switch a lamp, flip a relay, change how often a sensor reports — by asking.
It stays careful about it, which matters when something physical is about to happen in your house. It only does things your device is already set up to do — it does not invent new ones — and it works with the sensors connected to Chirp rather than gear that lives in someone else's app. (Bridge a Zigbee plug in over MQTT and it becomes an ordinary Chirp device the helper can switch.) It shows you what it is about to do and waits for you to say yes. And because a message to a sensor takes a moment and a sleeping device can miss one, it checks afterwards and tells you if it did not land, rather than just saying "done".
[→ Ask it to turn things on](/ai-assistant/let-ai-set-it-up.md)
***
**Control your home from ChatGPT or Claude**
And it isn't only the helper inside Chirp. Connect the AI app you already use — **ChatGPT**, **Claude**, or anything else that speaks MCP — sign in with your usual Chirp account, and it can switch things in your house too.
Same rules apply: only the things your devices are already set up to do, the same confirmation before anything happens, and the same check afterwards that the device actually got the message. The difference is where you are standing when you ask.
[→ Your own AI app](/api/mcp-server.md)
***
**What's New in the app**
Small things ship all the time and most of them go unnoticed. There is now a **What's New** panel in the app that walks you through what changed, one card at a time — so a release like this one does not pass you by.
***
**Smoothing things out**
Your subscription now shows the right dates for the current period. Adding a widget no longer nudges the widgets you had already arranged. On a phone, a long address no longer runs off the edge and hides the delete button. The label box on an alert threshold is wide enough to actually read what you typed, and the tooltip on sensor mapping now says what it means. Plus the usual round of quieter fixes underneath.
Flight Log. Release 3.7.0
3.7.0 is about letting your home out of the browser tab. A dashboard can now become a **link** — password-protected, yours to switch off — so the tablet in the hallway shows the house at a glance, and the person watering your plants next week can check on things without you handing over your account. When a new sensor goes quiet, the **Connection** tab finally *tells you why*, in words, with the next thing to try. Your sensors' setup codes get a proper **vault** instead of a drawer full of stickers. You can point your own AI app straight at your home. And adding a sensor no longer means squinting at 32 characters on the back of it — just **scan the code**. [app.chirpwireless.io](https://app.chirpwireless.io)
***
#### What's in This Release
* **Share your dashboard** — Turn any dashboard into a password-protected link. Pick **View** so people can look, or **Control** so they can actually switch things on. Revoke it whenever you like.
* **Connection diagnostics** — Your sensor's Connection tab now says whether data is arriving, what happens to it, and exactly what to check when nothing shows up.
* **Key Vault** — An encrypted spot for your sensors' EUI and AppKey codes, so you can find them again years later without the sticker.
* **MCP server** — Connect the AI app you already use — Claude Code, Claude Desktop, ChatGPT, Codex — to your home and ask it about your sensors.
* **Scan a QR code** — Point your phone or laptop camera at the sensor instead of typing a long code by hand.
* **Copy and move widgets** — Duplicate a widget you've already set up, or move it to a different dashboard.
* **Smoothing things out** — Layout, mobile, and navigation fixes across dashboards, sensors, and the activity log.
***
**Share your dashboard — a link you can hand out, and take back**
Until now, a dashboard was something only you and the people in your home could open. That's fine until you want the old tablet in the kitchen showing the house, or your neighbor is feeding the cat for a week, or your parents just want to see that the holiday home hasn't frozen.
Open a dashboard, tap its menu, and choose **Share dashboard**. Set a password, hit **Generate link**, and you have an address that opens the dashboard full screen — no Chirp account needed at the other end. Choose **View** and they can look but not touch. Choose **Control** and they can genuinely switch your devices on and off — so pick that one deliberately, and only for people you'd hand a key to.
The nice part is the off switch. **Change password** when the audience changes, **Regenerate link** to kill the old address, or **Revoke** to end it completely — the moment you do, the link stops working for everyone who has it.
[→ Sharing Your Dashboard](/dashboards/sharing-your-dashboard.md)
***
**Connection diagnostics — why isn't my sensor showing anything?**
It's the most deflating moment in a smart home: the sensor is on the wall, the batteries are in, you saved everything correctly — and the screen just says nothing. Until now you were left guessing whether it was the sensor, the range, the key you typed, or simply patience.
The **Connection** tab now answers it. At the top, a **reception status** in plain words: *Receiving & storing*, *Waiting for first data*, *Sending data — set up mapping to keep it*, or *Hasn't reported — device looks offline*. Below that, an **event feed** shows what happened to each message that came in, and a legend explains what every status actually means — including that **Skipped** is usually normal, not a fault.
Best of all, each unhappy state comes with a short **what to check** list — is it powered, is it in range of your gateway, does the AppKey match — and a button that takes you to the fix when the fix is in Chirp.
There's one answer in there that used to be impossible to find. A LoRaWAN sensor can only belong to one network at a time, so a second-hand find, a sensor that came with the house, or one that used to be on another app is still joined to *that* network — and adding it to Chirp won't change its mind. It'll sit on *Waiting for first data* forever while everything you can check looks right. Reset it, and it asks to join again.
[→ Connection Diagnostics](/devices/connection-diagnostics.md)
***
**Key Vault — stop keeping your keys on paper**
Every LoRaWAN sensor arrives with two codes on a tiny sticker: a DevEUI and an AppKey. You use them once, during setup, and then the sticker goes in a drawer, or the box goes in the recycling. Two years later the sensor needs re-pairing after a flat battery, and now you're on a ladder with a torch trying to read a sun-bleached label. A few sensors will hand the code back if you wire something to them; most won't, and either way it's nobody's idea of a Saturday.
**Key Vault** is an encrypted notebook in your home for exactly those pairs — one you don't have to go up a ladder to read. Save them straight from the sensor form with **Add to Vault** as you set a sensor up, or add them by hand later. Search by any part of a code to find one. It lives under the new **Records & Reports** section in the sidebar, next to your activity log, and who can open it is up to you.
[→ Key Vault](/reports/key-vault.md)
***
**Bring your own AI app**
Chirp already has a helper built in. Now you can also connect the AI app you use elsewhere — Claude Code, Claude Desktop, ChatGPT, Codex, or anything else that speaks MCP. Point it at your home's MCP address, sign in with your normal Chirp account in the browser — nothing to copy, no keys to paste — and you can ask it things like *which sensors have gone quiet?* or have it add a new sensor for you, right from your desktop. It only ever sees what your account can see. MCP is a shared standard, so as more apps adopt it, they'll work here too.
[→ MCP Server](/api/mcp-server.md) · [→ Your Home AI Helper](/ai-assistant.md)
***
**Two small things that save real annoyance**
Adding a sensor used to mean copying a 32-character code off a label without a single typo. Now there's **Scan QR code** — point your phone or laptop camera at the sensor and it fills itself in. And if you've spent time getting a widget looking just right, you can now **Duplicate** it for your other identical sensors, or **Move to dashboard** to send it somewhere it fits better.
[→ Adding Sensors](/devices/adding-sensors.md) · [→ Adding Widgets](/dashboards/adding-widgets.md)
***
**Smoothing things out**
This release also tidies up a handful of rough edges: dashboards now resize properly on a big monitor instead of scrambling their layout, adding a sensor works properly on a phone, a sensor's location can be changed or cleared rather than set once and stuck, the date picker on the activity log sits where you'd expect, and the gateway photo form now matches every other form in Chirp.
Flight Log. Release 3.6.0
3.6.0 is the one where Chirp goes **AI-first**. Your AI Helper grows up: ask it in plain English and it builds the automation, sets up the sensor, or creates the alert for you — no menus, no fiddly logic, always checking with you first. And your automations learn to *act*, not just warn: a new step lets a rule flip a switch on its own, so "if the basement gets damp, turn on the dehumidifier" happens by itself. This is **AIoT** — real intelligence built into your home. [app.chirpwireless.io](https://app.chirpwireless.io)
***
#### What's in This Release
* **An AI Helper that does the setup** — Ask it to add a sensor, build an automation, or set up an alert, and it does it for you — writing the logic, trying it out, and switching it on, always checking before anything permanent.
* **Automations that act on their own** — A new step lets an automation send a device a command by itself — turn on the dehumidifier, shut off the water, nudge the thermostat — the moment something happens.
* **Connect commands over the API** — The device-control calls are now part of the API, with new **Commands** permissions for your API keys.
* **iFrame widget** — Pin a live web page — a weather map, live traffic, the family calendar — right onto your dashboard, next to your sensors.
* **Tidier logs** — Your sensor's log now groups readings by the minute, and a little status dot in the header shows whether it's sending data right now.
* **Portuguese** — Chirp now speaks Portuguese.
* **Smoothing things out** — A handful of sign-in and app fixes to make everyday use steadier.
***
**Your home, now hands-on**
This is the big one. Years ago Chirp was one of the first to let you simply *ask* about your home's data — then we went quiet for a while, not because we'd stopped, but because we were building the engine underneath properly instead of rushing it out. 3.6.0 is the result: a Helper that genuinely does the work. Find it under **AI Chat** and just talk to it.
It answers from your real home — *"was the back door opened last night?"* — and it rolls up its sleeves: tell it "warn me if the nursery gets too warm" and it builds the automation, tries it, and turns it on; ask it to add your new leak sensor and it walks the whole thing through. Before anything big or permanent it stops and asks you to confirm, and once it's done it double-checks its own work.
The Helper runs on [SyntheticBrew](https://syntheticbrew.ai/), the AI agent runtime built by our team. Visit its public website for background on the runtime powering CHIRP's home assistant.
[→ Your Home AI Helper](/ai-assistant.md)
***
**Automations that act, not just alert**
Up to now an automation could *tell* you something was up; now it can *do* something about it. A new step lets an automation send a device one of its commands all by itself, the moment your conditions are met. The basement creeps past 70% humidity and the dehumidifier comes on without you lifting a finger. A leak sensor gets wet and the water shuts off on its own — then you get the alert, so you know it happened. Pick the device and the command, set each value as a fixed number or a little formula based on the reading, and your home starts looking after itself.
[→ When an Automation Runs a Command](/rules-engine/reference/automation-runs-a-command.md) · [→ Controlling Your Devices](/devices/commands.md)
***
**Connect commands over the API**
The calls that control a device are now part of the API, so your own scripts and tools can switch things on and off, not just read data. New **Commands** permissions on your API keys let you hand a tool exactly that and nothing more.
[→ REST API](/api/public-rest-api.md) · [→ API Keys](/settings/api-keys.md)
***
**A web page, right on your dashboard**
Some of the things you check every day aren't sensors at all — the weather, the traffic before the school run, the shared calendar, a parcel on its way. The new **iFrame** widget puts those web pages straight onto your dashboard, so they sit next to your temperature and door sensors instead of in yet another browser tab. Pick **iFrame** in the widget picker, paste the embed link the service gives you under *Share → Embed* — just the `https://` address, not the whole `
