Connect IQ App Development Discussion - Recent Threads

GPS Accuracy

DaveBrillhart — Sat, 01 Aug 2026 14:08:00 GMT

Cool!

Gemini: I have two Garmin devices. Here are the raw GPS coordinates from the same timestamps. How far apart are these?

Summary

Across this 4-second sample, the EPIX Pro Gen 2 and Edge 1050 were separated by an average of 1.29 meters (4.23 feet), converging slightly from 1.68 m down to 1.06 m. This level of delta is well within standard multi-band GNSS accuracy offset when mounting one device on the handlebar and wearing the other on the wrist.

Discussion about gear calculator data format for graph

bunny.hu — Fri, 15 Aug 2025 18:46:38 GMT

I nearly finish my gear calculator datafield. Yes, there is already some, but non of them know what I like.

One of my idea is save the calculated front and rear gear information into the FIT file for a late analize or whatever. But the graph can show only one number, and I am not sure, what is the best number. I ask the AI [emoticon:6d505171faa4497c85c5ca27290c555d] and it suggested that I thinking about earlier, so not better idea.

Let see my possible ideas for the numeric value:

Combined total speed gear like on a 2x12 system a number from 1-24. Maybe ordered by ratios or simple one chainring&cassette gears after the another.
Its good, when the calculation is jumping between the chainrings (the ratio is fit for more combination) there is no big jump here it ratio ordered but there is no visual information about when I change chainring.

Gear ratio Float number. Also nice continuos shifing even if the ratio jumping, but there no information at all about the real shifter position. Maybe a 3th part program can use it and count back to chainring&cassette position.

Combinated chainring&cassette position in one number. What I mean, the 101 means 1. chainring 1 cassette. 309 means 3. chainring 9. cassette. etc.

Combinated Ratio & chainring&cassette. Just like before but add the ratio like ratio*1000 . chainring&cassette. e.g.: 125.205 means 1.25 ratio , 2. chainring 5. cassette

What you thing about it? Maybe a better idea, or what is your favourite from them?

RE: Discussion about gear calculator data format for graph

9672100 — Sat, 01 Aug 2026 11:43:16 GMT

That’s a great point regarding data flexibility! Storing the pure gear ratio as the primary metric is definitely the most future-proof approach for cross-drivetrain comparisons and deeper post-ride analysis.

The nice thing about logging both having the gear ratio in its own dedicated graph while using the combined integer for quick visual inspection is that you get the best of both worlds: immediate readability on Garmin Connect's chart display without losing the raw underlying data for long-term tracking.

RE: Discussion about gear calculator data format for graph

9672100 — Sat, 01 Aug 2026 11:42:24 GMT

The nice thing about logging both—having the gear ratio in its own dedicated graph while using the combined integer for quick visual inspection—is that you get the best of both worlds: immediate readability on Garmin Connect's chart display without losing the raw underlying data for long-term tracking.

RE: Simulator problem

7333836 — Thu, 30 Jul 2026 05:52:16 GMT

Mystery solved—it was my mistake. Human stupidity really is limitless. Instead of "Run App", I had been using "Run Tests" the whole time.

I only figured it out after installing several different virtual machines with multiple SDK and Java versions.

Thanks a lot for your willingness to help, and I'm sorry for wasting your time.

Simulator problem

7333836 — Mon, 20 Jul 2026 12:02:52 GMT

Hello, I would need some advice regarding the simulator.

I used to create some watch faces in VS Code, but now I have a problem even with a new, clean project. When I try to start the simulation, the simulator opens, the selected watch briefly flashes on the screen, and then an empty simulator window remains open. Debug does not start, and no error message appears.

Can anyone advise me what might be wrong?

I have the latest version of VS Code, the latest Garmin SDK as well, and I have tried different Java versions, from Java 8 up to the latest one.

Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

5120468 — Mon, 01 Jun 2026 03:32:08 GMT

I am a long-term Garmin user (Fenix 6X Sapphire) and a scientist by profession. I am writing to formally request that Garmin consider implementing a Model Context Protocol (MCP) integration with Anthropic's Claude AI platform.

MCP is the open standard Anthropic uses to connect Claude to third-party data sources and applications. Several major platforms — including Google Drive, Gmail, and Microsoft 365 — have already implemented MCP connectors, enabling Claude to access and analyze user data in a secure, permissioned environment.

The case for a Garmin MCP integration is straightforward: Garmin Connect already collects one of the richest longitudinal physiological datasets available to consumers — Body Battery, HRV, sleep architecture, pulse ox, training load, stress scores, and activity history. What the platform currently lacks is deep interpretive analysis. Claude could provide exactly that: pattern recognition across multiple simultaneous data streams, plain-language explanation of trends, anomaly flagging, and contextual insights that go well beyond what dashboards alone can communicate.

This would be genuinely useful to a broad segment of your user base — endurance athletes, people managing chronic health conditions, and anyone engaged in serious long-term self-monitoring. It would also differentiate Garmin's ecosystem in a meaningful way, complementing rather than competing with Connect+.

Implementation would not require exposing raw data without consent. Standard OAuth permissioning, scoped API access, and user-controlled data sharing are all established patterns that other MCP integrations already follow.

I recognize this is a non-trivial development investment. I raise it because I believe it represents a genuinely high-value use case, and because user feedback is presumably one input into your roadmap prioritization.

Thank you for considering this request.

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

3328973 — Wed, 29 Jul 2026 18:28:15 GMT

Not sure how many people want to send their PHI through a 3P MCP... Garmin needs to own this.

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

juanitourquiza — Wed, 29 Jul 2026 17:55:34 GMT

Hey guys, now you can find the MCP server in the directory MCP Server

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

3328973 — Wed, 29 Jul 2026 14:43:13 GMT

This is a no-brainer. Built a project in Claude to manage nutrition and fitness. In order to get the fitness data, I need an MCP server for Garmin.

Because you guys are gating (maybe not deliberately - perhaps its in near-term roadmap?), I now have a convoluted config that has Garmin pushing data to Strava (I just signed up for an account with them because they're the only integration I found which you support that provides their own MCP). Here's the real burn... it's going to cost me $7 a month to Strava, literally only for the Garmin data. AND their MCP is in eligibility-based rollout, so I don't even have access to it yet.

Prioritize this above all else or you're going to lose people to other health and wellness platforms.

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

juanitourquiza — Wed, 29 Jul 2026 14:26:30 GMT

The same process of Taxuspt/garmin_mcp. I use it in the background inside pulsai.me, but you don't need extra configuration because you just need your MCP and this url the generate Pulsai. Please check the solution; I want your feedback

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

Vaclav Slajs — Wed, 29 Jul 2026 13:57:16 GMT

Hi Ilya,

Good news and bad news. The good news: emptying those OAuth fields was the right fix. I can see in the server logs that you got past the registration error and reached the sign-in page, which is exactly where you should be.

The bad news is what happened next, and it wasn't your fault or the connector's. Your seven sign-in attempts that afternoon all failed on Garmin's side — their login endpoints were returning rate-limit errors and CAPTCHA challenges to us for the whole day. Roughly two thirds of all sign-in attempts to this server failed that day, so you happened to pick the worst possible afternoon to try.

Garmin has since gone back to normal — no CAPTCHA challenges at all today. Please give it one more try; the same credentials should work now.

One practical note: if a sign-in does fail, don't retry immediately in a row. Garmin tightens the screws per account when it sees repeated attempts, so several tries in a few minutes makes the next one more likely to fail, not less. One attempt, and if it fails, leave it 15–30 minutes.

If it fails again, tell me roughly when you tried and I'll look up exactly what Garmin returned — the logs show me the reason, so I don't have to guess.

Sorry for the run-around. You found a genuine problem in the error message, and it's fixed for everyone now.

Best,
Vaclav

RE: Help with ERA errors

9665893 — Wed, 29 Jul 2026 08:37:08 GMT

That's a helpful explanation. ERA errors can be difficult to interpret at first, so it's great to have some context. Thanks for taking the time to share this.

Help with ERA errors

Gumix — Mon, 19 Jan 2026 22:16:18 GMT

Hi,

I've got the errors like "Invalid Value" and "System Error" on the code with simple arithmetic operations. What could be causing this?

Source code is here.

Reports:

[embed:dc8ab71f-3b98-42d9-b0f6-e21e02a0f8e2:0e0a8f17-48ea-4524-9e66-e92df76f2412:type=text&text=Error%20Name%3A%20Invalid%20Value%0AOccurrences%3A%202%0AFirst%20Occurrence%3A%202026-01-17%0ALast%20Occurrence%3A%202026-01-18%0ADevices%3A%0AVenu%C2%AE%20Sq%202%3A%202.69%0A%C2%A0epix%C2%A0Pro%20%28Gen%202%29%2051mm%20%2F%20D2%C2%A0Mach%201%20Pro%20%2F%20tactix%C2%AE%207%20%E2%80%93%20AMOLED%20Edition%3A%2023.48%0AApp%20Versions%3A%202.0%0ALanguages%3A%20eng%2C%20ita%0ABacktrace%3A%0AIV22LView.onLayout%3A91]

[embed:dc8ab71f-3b98-42d9-b0f6-e21e02a0f8e2:ddc0ee00-2bfe-45fe-9025-9c7781784e3d:type=text&text=Error%20Name%3A%20System%20Error%0AOccurrences%3A%202%0AFirst%20Occurrence%3A%202026-01-16%0ALast%20Occurrence%3A%202026-01-19%0ADevices%3A%0AVenu%C2%AE%20Sq%202%20Music%3A%202.69%0Av%C3%ADvoactive%C2%AE%205%3A%2017.01%0AApp%20Versions%3A%202.0%0ALanguages%3A%20eng%2C%20por%0ABacktrace%3A%0AIV22LView.onLayout%3A92]

[embed:dc8ab71f-3b98-42d9-b0f6-e21e02a0f8e2:d3188d13-be5e-49d0-b52f-767f041f74ea:type=text&text=Error%20Name%3A%20Invalid%20Value%0AOccurrences%3A%201%0AFirst%20Occurrence%3A%202026-01-16%0ALast%20Occurrence%3A%202026-01-16%0ADevices%3A%0Af%C4%93nix%C2%AE%208%2047mm%20%2F%2051mm%20%2F%20tactix%C2%AE%208%2047mm%20%2F%2051mm%20%2F%20quatix%C2%AE%208%2047mm%20%2F%2051mm%3A%2020.24%0AApp%20Versions%3A%202.0%0ALanguages%3A%20eng%0ABacktrace%3A%0AIV22LView.onLayout%3A93]

[embed:dc8ab71f-3b98-42d9-b0f6-e21e02a0f8e2:4ae6f575-2385-48c4-8314-4ab1dc2a9758:type=text&text=Error%20Name%3A%20System%20Error%0AOccurrences%3A%202%0AFirst%20Occurrence%3A%202026-01-16%0ALast%20Occurrence%3A%202026-01-18%0ADevices%3A%0Av%C3%ADvoactive%C2%AE%205%3A%2017.01%0AForerunner%C2%AE%20970%3A%2016.14%0AApp%20Versions%3A%202.0%0ALanguages%3A%20eng%0ABacktrace%3A%0AIV22LView.onLayout%3A93]

Log prints on Watches

DaveBrillhart — Tue, 28 Jul 2026 16:26:19 GMT

Interesting inconsistency.... Just FYI. Easy workaround putting it in the STOP method.

For my EDGE devices, I can write out a final set of diagnostics to a LOG file in the onTimerReset() call, once you hit STOP and then either SAVE or DISCARD.

That wasn't happening on my EPIX PRO GEN2 watch. So I had to do that in onTimerStop(). Nothing was written to the LOG from the println statements in onTimerReset().

RE: Log prints on Watches

jim_m_58 — Tue, 28 Jul 2026 17:32:49 GMT

It's not the logging. It's that onTimerReset isn't called on all devices.

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

ivomilkovic — Tue, 28 Jul 2026 16:44:39 GMT

this is really needed

RE: Unexpected BLE connection problem

NVelev — Tue, 28 Jul 2026 16:20:25 GMT

Of course your're right about that and it is clearly stated in the API documentation:

getServices() as BluetoothLowEnergy.Iterator

Retrieves an Iterator over the services provided by the device.

This will only provide Services that have been registered using registerProfile()

However the service has already been properly registered with its UUID and respective characteristics in the same way as I've done it for other applications, where executing getServices() on a search result did return non-empty results.

Also, as I've mentioned, after I pair directly I can use this service and read the values, which wouldn't be possible unless the service has been registered before that..

Apparently there is some form of incompatibility with this particular device.

Unexpected BLE connection problem

NVelev — Fri, 03 Jul 2026 05:55:02 GMT

I'm working on an app that uses data from a remote pressure sensor and I believe I got stuck because of a rather unexpected problem. My Garmin Edge 1030 device is not being able to discover the sensor during BLE scan.

As a background, I want to state that I've successfully used connections to several other BLE enabled external devices in other CIQ apps on the same Edge 1030 device. So it is not that I'm missing out how to do it in principle.

The problem now is that the list of results that are returned to the BleDelegate after initiating a BLE scan on the Edge 1030 device just don't contain the sensor I'm playing with.

All sorts of other devices can be seen there, but never this one.

Without discovering it via the built-in scan I cannot connect to it because the "pairDevice()" function accepts as a parameter a ScanResult object and a ScanResult object cannot be instantiated manually. It has to be created by the BLE scan..

Can anyone come up with a suggestion about why the sensor is not "visible" for the BLE scan that the Edge 1030 performs, while in the same time it is readily visible when I do a scan on my Android phone?

Thank you in advance!

Nikolay

RE: Unexpected BLE connection problem

jim_m_58 — Tue, 28 Jul 2026 14:24:10 GMT

CIQ doesn't discover services and characteristics. They must be defined in the profile manager (see ProfileManager.mc in the nordicThingy52 sample in the SDK).

Also note, you can define at most 3 services there. Too many will cause an exception. (registerProfile)

RE: Unexpected BLE connection problem

NVelev — Tue, 28 Jul 2026 14:11:53 GMT

Good question, but how do I know whether there are BLE advertising modes / services that are not supported by CIQ?

I could not solve the puzzle about a straightforward way to find this device with a CIQ BLE search.

The name is null in search results, despite that on my phone I can see a valid name.

The getServices() returns an empty iterator, despite that there is a service defined in the sensor's BLE configuration.

Since I know the MAC address of the specific unit I just go through the search results checking the hasAddress() function until there is a match. Then I pair directly and subscribe for the notification events of the known sensor's service and I can then read the measured values without a problem.

However, why getServices() is not able to discover that there is a configured service in the device, is something that I didn't figure out.

App communication with watch

Greg keene — Tue, 28 Jul 2026 00:37:05 GMT

TITLE:
iOS companion: device returns from getConnectedDevices but reports NOT_PAIRED, getApplicationInfo says NOT_INSTALLED, and every Communications.transmit() hits onError

BOARD:
Connect IQ > Connect IQ App Development Discussion

SETUP

Watch: Forerunner 245 Music (fr245m), firmware current
SDK: Connect IQ 9.2.0 on Windows, building with the monkeyc CLI (the VS Code
extension is unusable for me - CIQQA-3791)
Watch app: device app (type="watch-app"), manifest declares Positioning and
Communications permissions and lists fr245m
Companion app: iOS, React Native / Expo, using the community wrapper
react-native-connect-iq-mobile-sdk over the official iOS Mobile SDK
Distribution: the watch app is uploaded to the Connect IQ Store as a beta app
and installed on the watch through the Connect IQ mobile app (not sideloaded).
The manifest appId and the store appId match.

WHAT WORKS

The AppDelegate openURL hook is wired up and confirmed firing. Garmin Connect
returns a well-formed callback, for example:

myscheme://device-select-resp?ciqApp=Connect&ciqBundle=com.garmin.connect.mobile
&ciqSdkVersion=10000&d0ID=<uuid>&d0Model=Forerunner%20245%20Music
&d0Name=Forerunner%20245%20Music&d0PartNumber=006-B3077-00

getConnectedDevices() returns exactly one device with the correct friendlyName
("Forerunner 245 Music").
setDevice() on that returned object succeeds.
registerForAppMessages(appId) resolves without error.
The watch and phone are paired and connected normally. Garmin Connect Mobile and
the Connect IQ app both show the watch as Connected, syncing works, and the
watch's "Find My Phone" successfully makes the phone ring - so the BLE link and
the device-level pairing are clearly healthy.

WHAT FAILS

1. The device status callback reports NOT_PAIRED for that device ID (the same
UUID that came back in the d0ID param).
2. getApplicationInfo(appId) reports {"status":"NOT_INSTALLED","version":"0"},
even though the app is installed via the store and visibly runs on the watch.
3. On the watch, Communications.transmit(payload, null, listener) always lands in
ConnectionListener.onError(). Confirmed on-device by drawing the result on the
watch screen, since System.println isn't visible on real hardware.
4. No message is ever delivered to the companion app's message listener.

One other oddity: in the Connect IQ mobile app, "My Device Apps" showed a count
of 1 for this watch while rendering an empty list.

CONTROL TEST - GARMIN'S OWN COMM SAMPLE FAILS THE SAME WAY

To check whether my watch code was at fault, I built the SDK's own Comm sample
(added fr245m to its products list, no other changes), sideloaded it, and pointed
my companion app's registerForAppMessages at the Comm sample's own app id
(a3421feed289106a538cb9547ab12095).

I sent "hello world" from the Comm sample's Send menu on the watch. Nothing
arrived at the companion app. getApplicationInfo for the Comm sample's app id
also returns NOT_INSTALLED, and the device still reports NOT_PAIRED.

So Garmin's own reference watch app behaves identically to mine, which seems to
rule out my watch-side code.

I can't run Garmin's iOS example companion app to complete the isolation - no Mac
or Xcode available - which is why I'm asking here.

ALREADY RULED OUT

- Registering for app messages with an empty-string appId - no effect
- Sideload vs store install - same behavior either way
- appId mismatch - manifest appId and store appId now match; also tried the phone
side registering an older appId to match an older build
- App id format - tried both a dashed GUID and a conventional undashed 32-hex id
(including Garmin's own sample id); identical results
- Handing setDevice() a hand-constructed device object vs. the object returned by
getConnectedDevices() - same result either way. Note the returned object's
status field comes back undefined through the wrapper.
- Call ordering - re-registered the device-event and app-message listeners
immediately after setDevice(), in case registerForDeviceEvents was being called
before a device existed. No change.
- Companion app backgrounded - the app is open and in the foreground during tests
- Bluetooth range or connection - the watch is connected and syncing throughout

QUESTIONS

What causes a device to report NOT_PAIRED when it was just returned by the device
selection flow, and is that status the reason transmit() is refused? My working
assumption is that everything else here (NOT_INSTALLED, the refused transmits,
the empty app list) is downstream of that one status, since an invalid device
can't be queried about apps or routed messages.

Is there a registration step between setDevice() and registerForAppMessages()
that I'm missing on iOS?

Is getApplicationInfo expected to return NOT_INSTALLED for a beta store app that
is installed on the device? If beta apps stay permanently "pending", does that
block app-message routing?

Happy to provide logs or put together a minimal repro.

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

juanitourquiza — Mon, 27 Jul 2026 16:12:10 GMT

Hey, you can try this option Pulsai. I'm working on this product every day

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

1461130 — Mon, 27 Jul 2026 16:03:54 GMT

Thank you for the prompt and detailed answer, Vaclav!

With your hint, I moved one step further, but there I received the message "Garmin sign-in timed out — the service may be rate-limiting new sign-ins. Please wait a moment and try again." Though this may be due to my numerous experiments with different services today 🙂

Well, meanwhile, I managed to set up "Taxuspt\garmin_mcp", will see how it works. But definitely your bridge will be an easier solution for people like me, who start to be nervous every time when they need to open the console 😃

RE: Feature Request: MCP Integration to Enable Claude AI Access to Garmin Connect Data: To the Garmin Product and Developer Teams,

Vaclav Slajs — Mon, 27 Jul 2026 13:47:21 GMT

Hi Ilya,

Thanks for the detailed report — you found a real problem, and the error message was actively misleading you. Removing and re-adding the connector genuinely doesn't help, because you were never in the state that message describes.

Here's what's happening. When you add the connector, Claude shows two optional fields, OAuth Client ID and OAuth Client Secret. If either one is filled in, Claude stops registering itself with the server automatically and sends your value instead — and the server has no such client, so it refuses. Those fields aren't your account details; the server issues its own credentials behind the scenes.

To fix it:

1. Remove the connector (Settings → Connectors → Remove).
2. Add it again with only the URL: missingmcp.com/.../mcp
3. Leave OAuth Client ID and OAuth Client Secret completely blank.
4. Claude will open a sign-in page — that's where your Garmin email and password go.

Incognito won't change anything here, so no need to bother with it.

I've since made the server explain this properly instead of sending you in a circle, and the setup instructions on the site now say to paste the URL and nothing else. You weren't the only one hitting it — six people ran into the same wall last week, so your message fixed it for all of them.

One thing that would help me confirm the diagnosis: were those OAuth fields in fact filled in on your side? I'm fairly confident from the server logs, but I haven't seen your setup.

Best,
Vaclav