/ FORMATS & INTEGRATIONS

Capture once. Export anywhere.

Every trajectory is recorded a single time and converted on demand — into the on-disk format your tools expect and the destination your stack already uses. Open standards in, open standards out, zero lock-in.

12+export formats
20+connections
0lock-in

/ FORMAT EXPLORER

Pick a format. See exactly what you get.

MCAP
A modern, indexed container purpose-built for robotics — append-only, schema-aware, and fast to seek across many topics.
.mcap
BEST FORMultimodal logs at scale
TYPE
Container
SCHEMA
Self-describing
RANDOM ACCESS
Indexed
COMPRESSION
zstd / lz4
terminal
$ telemanual export --to mcap dexterous-pick-v2

/ DATA CONNECTIONS

Sync to the destinations you already run.

Push to object storage, warehouses, experiment trackers, streaming buses, or pull through our APIs. We handle schema, sync, and retries.

Amazon S3object store
Google Cloudobject store
Azure Blobobject store
Snowflakewarehouse
Databrickslakehouse
Hugging Facemodel hub
Weights & Biasesexperiment tracker
Apache Kafkastreaming bus
REST APIpull / push
gRPCstreaming RPC

/ HOW EXPORT WORKS

One capture, any number of destinations.

01

Capture once

Telemanual records a canonical, lossless, time-aligned trajectory across every sensor.

02

Convert on demand

Choose a format and we transcode — schema, compression, and indexing handled.

03

Sync anywhere

Land it in your bucket, warehouse, or hub, or pull it through the API.

No lock-in, ever

Open, documented formats you can read with standard tools. Your data is portable by default.

Reproducible

Versioned, content-addressed exports mean a dataset is the same dataset everywhere it lands.

Pipeline-native

Drop straight into ROS, PyTorch, TensorFlow, Spark, or your warehouse — no glue code.

/ FAQ

Formats & integrations, answered.

See your data land in your stack.

Start a pilot and watch your first trajectories arrive in the format and bucket you choose — within days.