Roboto SDK

Roboto is the analytics engine for Physical AI: ingest, search, and analyze your robotics data at scale, and put AI agents to work on it 🤖

This package is the official Python SDK for Roboto. The roboto command line utility is distributed separately as standalone binaries (see CLI below).

If this is your first time using Roboto, start with the docs and the core concepts.

Why Roboto?

Most robotics teams start with manual review: visualizing logs and replaying data. But that approach alone doesn't scale; your fleet generates more data than your team can ever review. Roboto helps you go from raw logs to root cause 🚀

Ingest logs in every major robotics format, query data across your fleet, and define custom actions to post-process it: identify events, generate KPIs, and more.

You can also let AI agents do the analysis: they search your data, dig into signals, summarize and triage datasets, and detect events, whether in the Roboto web app, through this SDK, or from the AI tools you already use.

See below for supported data formats, installation instructions, and getting started examples.

Connect Your AI Tools

The hosted Roboto MCP server brings Roboto to the AI you already work in. Connect it once, and Claude Code, Claude Desktop, Codex, Cursor, VS Code, or any other client that speaks the Model Context Protocol can search your Roboto data, explore datasets and files, and analyze topic data mid-conversation:

claude mcp add --transport http roboto https://mcp.roboto.ai/mcp

You authenticate as yourself — sign in with your browser or use a personal access token; every tool call runs with your own permissions.
Read-only by design — a curated set of search, retrieval, and analysis tools; nothing exposed over MCP can modify your data.
All of your organizations, one connection — if you belong to several orgs, ask the AI to call whoami and set_active_org to switch between them.

See Use the Roboto MCP Server for setup instructions per client, and the Roboto MCP Server docs for everything the AI can do once connected.

Data Formats

Roboto ingests the following formats, each with a corresponding action in the Action Hub.

Format	Extensions	Status	Action
ROS 2	`.mcap`, `.db3`	✅	`ros_ingestion`
ROS 1	`.bag`	✅	`ros_ingestion`
PX4	`.ulg`	✅	`ulog_ingestion`
Parquet	`.parquet`	✅	`parquet_ingestion`
CSV	`.csv`	✅	`csv_ingestion`
ArduPilot	`.bin`, `.log`, `.tlog`	✅	`ardupilot_ingestion`
Video	`.mp4`, `.avi`, `.mkv`	✅	`video_ingestion`
Journal	`.log`	✅	`journal_log_ingestion`

Roboto can also support custom formats. Reach out to discuss your use case.

Install Roboto

To use the Roboto SDK or CLI:

Sign up at app.roboto.ai and create an access token (docs)
Save your access token to ~/.roboto/config.json

Python

The Roboto Python SDK is available on PyPI.

Requirements: Python 3.10+

Installation:

pip install roboto

[!TIP]

If pip install roboto fails or behaves unexpectedly, the two most common causes are a Python version below 3.10 and dependency conflicts from installing into your system Python or an existing environment.

To rule out both, confirm your Python version meets the requirement:

python3 -c "import sys; assert sys.version_info >= (3, 10), f'Python 3.10+ required, found {sys.version}'"

Create and activate a virtual environment:

# macOS/Linux
python3 -m venv .venv && source .venv/bin/activate

# Windows
python -m venv .venv && .venv\Scripts\activate

Then re-run pip install roboto.

Optional features:

pip install roboto installs the core SDK. Some functionality depends on heavier third-party libraries that ship as optional extras, so you install them only when you need them:

Extra	Install	Enables
`analytics`	`pip install 'roboto[analytics]'`	Searching for similar signals (`roboto.analytics.find_similar_signals`) and reading topic data into pandas DataFrames (`Topic.get_data_as_df`, `MessagePath.get_data_as_df`). Adds `fsspec`, `numpy`, `orjson`, `pandas`, `pyarrow`, and `stumpy`.
`ingestion`	`pip install 'roboto[ingestion]'`	Parsing and writing topic data as Parquet, e.g. when building ingestion actions. Adds `pandas` and `pyarrow`.
`video`	`pip install 'roboto[video]'`	Extracting frames from video files (`roboto.experimental.video`). Adds `av`, `numpy`, and `pillow`.

Without the matching extra, these code paths raise an ImportError naming the extra to install. For example, calling find_similar_signals without the analytics extra raises:

ImportError: Missing optional dependency 'stumpy'. Re-install roboto using pip or conda with 'roboto[analytics]' to install a compatible version.

The examples extra provides the tooling needed to run the SDK's example notebooks, namely ipython, jupyter, matplotlib, and pillow. Use pip install 'roboto[examples]' before trying out the included examples.

[!IMPORTANT]

Quote the package spec when installing an extra. Zsh (the default shell on macOS) treats the square brackets as a glob pattern, so an unquoted pip install roboto[analytics] fails with zsh: no matches found: roboto[analytics]. Quoting the spec (pip install 'roboto[analytics]') avoids the error.

Authentication (required):

The SDK uses the access token you saved above; if you haven't created one yet, see Setting up programmatic access. Verify your credentials are configured correctly:

python -m roboto.cli users whoami

See the complete SDK documentation.

CLI

To use Roboto from the command line without the Python SDK, install the standalone CLI.

Pre-built binaries for every version are on the releases page of this package. We build for Linux (aarch64, x86_64), macOS (aarch64, x86_64), and Windows (x86_64). See per-platform installation instructions below.

The CLI provides the roboto command line utility. List available commands with roboto -h, or see the complete CLI reference documentation.

Linux

Go to the latest release page for this package
(apt) Download the relevant roboto .deb file for your platform
- e.g. roboto-linux-x86_64_0.9.2.deb (not a roboto-agent release)
- Double-click the downloaded .deb file to install it with apt
(non-apt) Download the relevant roboto file for your platform
- e.g. roboto-linux-x86_64 (not a roboto-agent release)
- Move the downloaded file to /usr/local/bin or another directory on your PATH

Coming soon: direct apt-get install support

macOS

Install with the Homebrew package manager:

brew install roboto-ai/tap/roboto

Or download the relevant Mac binary, e.g. roboto-macos-aarch64, from the latest release page.

Windows

Go to the latest release page for this package
Download the roboto-windows-x86_64.exe file
Move the downloaded .exe to a folder on your PATH, like C:\Program Files\

Upgrade CLI

The CLI automatically checks for updates and notifies you when a new version is available.

Homebrew users can upgrade by running brew upgrade roboto. If you installed a .deb or a standalone executable, download the latest version and replace the old one.

Getting Started

Use the CLI for quick tasks: creating datasets, uploading or downloading files, and running actions. Use the Python SDK for full platform coverage, data analysis, and integration with your other tools.

CLI Example

The example below creates a new dataset and uploads a file to it. In a Python environment with the SDK installed, the same commands are available via python -m roboto.cli.

> roboto datasets create --tag boston
{
  "administrator": "Roboto",
  "created": "2024-09-25T22:22:48.271387Z",
  "created_by": "benji@roboto.ai",
  "dataset_id": "ds_9ggdi910gntp",
  ...
  "tags": [
    "boston"
  ]
}

> roboto datasets upload-files -d ds_9ggdi910gntp -p scene57.bag
100.0%|█████████████████████████ | 58.9M/58.9M | 2.62MB/s | 00:23 | Src: 1 file

Python Example

The example below accesses topic data from an ingested ROS bag file:

from roboto import Dataset

ds = Dataset.from_id("ds_9ggdi910gntp")
bag = ds.get_file_by_path("scene57.bag")
steering_topic = bag.get_topic("/vehicle_monitor/steering")

steering_data = steering_topic.get_data(
    start_time="1714513576", # "<sec>.<nsec>" since epoch
    end_time="1714513590",
)

You can also create events:

from roboto import Event

Event.create(
  start_time="1714513580", # "<sec>.<nsec>" since epoch
  end_time="1714513590", 
  name="Fast Turn",
  topic_ids=[steering_topic.topic_id]
)

Or search for logs matching metadata and statistics with RoboQL:

from roboto import query, RobotoSearch
roboto_search = RobotoSearch(query.client.QueryClient())

query = '''
dataset.tags CONTAINS 'boston' AND
topics[0].msgpaths[/vehicle_monitor/vehicle_speed.data].max > 20
'''

results = roboto_search.find_files(query)

Or put an AI agent to work on your data and stream its findings as it goes:

from roboto.ai import AgentThread
from roboto.ai.agent_thread import AgentTextDeltaEvent

thread = AgentThread.start(
    "Look at dataset ds_9ggdi910gntp and check /vehicle_monitor/steering for anomalies."
)

for event in thread.events():
    if isinstance(event, AgentTextDeltaEvent):
        print(event.text, end="", flush=True)

The same agents are available in the web app's AI Chat and on the command line via roboto chat start.

See the notebooks directory for complete examples!

Learn More

For more information, check out:

Contact

Email us at info@roboto.ai or join our community Discord server.

Metadata

Release files for roboto 0.55.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for roboto 0.55.0
File	Size	Uploaded
roboto-0.55.0.tar.gz	533.4 kB	Sep 1, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for roboto 0.55.0
File	Interpreter	ABI	Platform	Reset
roboto-0.55.0-py3-none-any.whl 741.5 kB Sep 1, 2026	Python 3	none	any	Details

Total release size: 1.3 MB

Release files / roboto-0.55.0.tar.gz

Download URL	roboto-0.55.0.tar.gz
Size	533.4 kB
Tags	Source
SHA-256 checksum How to use checksums	`5e44ed98af51d79085d5638f914fb3b142562f835bf5862e93af5ebe58a1e605`
BLAKE2b-256 checksum How to use checksums	`170a72e5b458e480165777ae88e725d4c6448b891864ab18cf7df53fcb73747d`
Upload date	Sep 1, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/roboto-ai/roboto-python-sdk
Publishing commit: github.com/roboto-ai/roboto-python-sdk/tree/d8c6d5bc3e82f160f0ac427b71be9a77e41bc713

Workflow

Publishing configuration: github.com/roboto-ai/roboto-python-sdk/blob/d8c6d5bc3e82f160f0ac427b71be9a77e41bc713/.github/workflows/sdk.yml
Publishing logs: github.com/roboto-ai/roboto-python-sdk/actions/runs/33564674265/attempts/1

Artifact

Subject: roboto-0.55.0.tar.gz
SHA-256 checksum: 5e44ed98af51d79085d5638f914fb3b142562f835bf5862e93af5ebe58a1e605 Verifying attestations

Release files / roboto-0.55.0-py3-none-any.whl

Download URL	roboto-0.55.0-py3-none-any.whl
Size	741.5 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`a3c9c6691a9ea2d15bdf8f3dc457b659f8b194616c4a56b1fe99d6093ddb0eb6`
BLAKE2b-256 checksum How to use checksums	`c97482cfbc7acac9c748850e4fc48a9b9d00b634177e73d4e36c235f3638e0cf`
Upload date	Sep 1, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/roboto-ai/roboto-python-sdk
Publishing commit: github.com/roboto-ai/roboto-python-sdk/tree/d8c6d5bc3e82f160f0ac427b71be9a77e41bc713

Workflow

Publishing configuration: github.com/roboto-ai/roboto-python-sdk/blob/d8c6d5bc3e82f160f0ac427b71be9a77e41bc713/.github/workflows/sdk.yml
Publishing logs: github.com/roboto-ai/roboto-python-sdk/actions/runs/33564674265/attempts/1

Artifact

Subject: roboto-0.55.0-py3-none-any.whl
SHA-256 checksum: a3c9c6691a9ea2d15bdf8f3dc457b659f8b194616c4a56b1fe99d6093ddb0eb6 Verifying attestations