What Fleet Telemetry actually is

The short version

Tesla's developer documentation describes Fleet Telemetry as a way for vehicles to stream data straight to a server run by the app's developer, so the app no longer has to poll the car's vehicle_data endpoint. Tesla gives the reason in the same breath: polling wakes cars that would otherwise sleep and costs them battery, while a stream only flows when the car has something to report.

Tesla is also clear that the stream is not an exact copy of vehicle_data. It carries the fields an app subscribes to, at the intervals the app asks for, and nothing else. What an app shows you is therefore a choice its developer made, and one worth asking about.

How a car starts streaming

Three things have to be true. The car needs firmware that supports it: Tesla lists 2024.26 or later for the current, signed form of the configuration. The owner has to pair the app's virtual key with the car, through a tesla.com link that opens the Tesla app. And the app has to send the car a configuration through Tesla's fleet_telemetry_config endpoint, signed through Tesla's vehicle-command proxy.

That configuration names the server, the fields and an interval for each. Tesla cannot alter it on the way, the car keeps it, and it is removed if the owner revokes the app's access. A car accepts configurations from only a few third-party apps at a time, and Tesla documents the reasons it can refuse one: no virtual key, unsupported hardware or firmware, or too many configurations already.

Removing the app's key on the car's Locks screen stops both streaming and commands at once, according to Tesla. That is the owner's off switch, and it does not depend on the app behaving.

Some older Model S and Model X cannot stream at all. Tesla's own pages disagree about exactly which: one names cars from before 2018 without the infotainment upgrade, another names cars from before 2021. For an early S or X, the honest test is to connect it and see what the car answers. Tesla has also announced that a Model S or X with the Intel Atom computer on firmware 2025.20 or later can stream once the owner turns on third-party data streaming on the car's Safety screen, without a virtual key.

What the interval really means

Every field in a configuration carries an interval in seconds, and it is easy to misread. It is not a polling rate. Tesla's documentation says a field is sent only when two things are both true: the interval has passed since it was last sent, and its value has changed. The car groups records into half-second batches before sending them.

So a car parked on a driveway with a one-second interval on speed sends nothing about speed, because its speed is not changing. A stream is cheap when nothing happens and detailed when something does: the opposite of a timer.

Newer firmware adds refinements: a minimum change before a value counts as changed, and an option to resend a value now and then even if it has not moved. If the car loses its connection it buffers messages and sends them when it is back; if the server cannot be reached, the car retries, waiting no more than 30 seconds between attempts.

Who pays for it

Tesla bills the app's developer, not the owner, and it prices each kind of traffic differently. On Tesla's developer site, one dollar buys 150,000 streamed signals, 1,000 commands, 500 data requests or 50 wakes. A single wake therefore costs as much as 3,000 streamed signals, and that ratio is the economic case for streaming.

Tesla also sets per-car limits on requests, commands and wakes, and its billing documentation calls frequent waking a sign of improper app design. An app built on the stream has no reason to wake your car; an app built on polling has to decide how often it can afford to.

How DashM8 uses it

DashM8 reads your car through the stream and nothing else. When you connect, we send your car its configuration and show you what the car answered, so a car that cannot stream is something you find out on the free plan. Our backend refuses to call wake_up or vehicle_data at all, and a command sent to a sleeping car is refused with a message asking you to wake the car yourself.

The intervals we ask for depend on your plan. On Pro, no field is requested more often than every few seconds. On the free plan most fields are held to once a minute, with Sentry Mode at ten seconds and doors and locks at thirty. Because sending is change-driven, these are ceilings: a parked car sends almost nothing either way.

What we cannot see is anything between the values the car chooses to send. Every figure in DashM8 carries the time it was last heard, and a field that has gone quiet says so rather than pretending to be live.

Sources

Every source below was checked on 7 October 2026. Prices, tariffs and Tesla's pages change; check the source before you rely on a figure.

1. Tesla Developer: Fleet Telemetry
2. Tesla Developer: Vehicle Endpoints
3. Tesla Developer: Virtual Keys: Overview
4. Tesla Developer: Announcements
5. Tesla Developer: Billing and Limits
6. Tesla Developer: Tesla Developer (pricing)
7. Tesla on GitHub: teslamotors/fleet-telemetry