PokeBot’s RoboAPI Lets Developers Try New Tasks on Their Own Robot Arms, With No Training Step
News originally published by
GlobeNewswireIn PokeBot’s evaluation of nearly 1,000 unseen objects and nearly 10,000 unseen tasks, Poke-Aurora achieved success rates of 74.2% across three task categories.
Beijing, China, Nov. 17, 2026 (GLOBE NEWSWIRE) -- Getting a robot arm to handle a task it wasn’t built for has typically meant collecting examples, training a model, testing it and repeating, a cycle that can take weeks. Last month, PokeBot, an embodied AI company that develops intelligent systems that understand and act in the physical world, launched RoboAPI, an interface that lets developers skip it. Connect a compatible robot arm with a wrist-mounted camera, give it an instruction in plain language, and watch how it handles the task in your own space.
- Watch robotics researchers put RoboAPI’s plug-and-play setup to the test in their own labs.
From plug-in to first try
RoboAPI is the interface, and Poke-Aurora is the model behind it. Developers and researchers get no-cost access to Poke-Aurora, PokeBot’s world-action model, with no data collection or fine-tuning required. Step-by-step installation takes about two hours, and the guide is available here.
Developers don't need to collect demonstration data, fine-tune a model or build a custom training pipeline before trying new tasks. They can connect compatible robot hardware, give Poke-Aurora a new instruction and see how it responds to unfamiliar objects and surroundings.
“Users can tell a robot arm what to do in plain language,” said Huazhe Xu, founder of PokeBot and an assistant professor at Tsinghua University’s Institute for Interdisciplinary Information Sciences. “The model looks at what the robot sees, figures out the instruction, and chooses its next move, so developers can watch how it performs on real hardware.”
A model that predicts before it acts
A world-action model predicts what will happen next, then chooses the robot’s action. Poke-Aurora was built to handle objects and settings it hasn’t encountered, rather than only the controlled setups many robot demos rely on. Because of this, developers can test how a robot responds to new instructions and changing surroundings without building a custom training pipeline first.
In PokeBot’s evaluation of nearly 1,000 unseen objects and nearly 10,000 unseen tasks, Poke-Aurora achieved success rates of 87.7% in Environment Interaction, 86.0% in Transfer & Organization, and 54.1% in Pose & Shape Alignment.
What developers could build toward
Developers and researchers could use RoboAPI to explore everyday household applications. Longer term, the same approach could support tasks like tidying, sorting items or putting dishes away, though these are examples of where the technology is headed, not tested results.
Developers and researchers can use RoboAPI to explore how Poke-Aurora handles practical tasks, such as wiping up a spill with a towel or standing a fallen bottle upright. Longer term, the same approach could support tasks like tidying, sorting items or putting dishes away, though these are examples of where the technology is headed, not tested results.
“Building smart robots is only part of the challenge,” said Xu. “Developers also need to see how robot AI works outside of a demo. With RoboAPI, they can try Poke-Aurora on real robots and see for themselves.”
Availability and resources
PokeBot launched RoboAPI on Oct. 30 with a technical paper and product demo. It is in technical preview, available to eligible users for noncommercial use, and works with compatible arms such as Universal Robots and Franka models, per the public package.
Developers can register with an email address at PokeBot’s official platform.
- Technical paper and updates: https://pokebot.com/ai
- Apply for early access: https://platform.pokebot.com/login
- API documentation: https://developers.pokebot.com/docs/roboapi
About PokeBot
PokeBot is an embodied AI company developing intelligent systems that enable robots to perceive, understand, and act in the physical world. Through its work on general-
purpose embodied AI models, the company aims to advance adaptable robot
intelligence and bring capable robots closer to everyday applications.
Learn more at pokebot.com.

Silver Yang media@silverbulletstrategy.com
Legal Disclaimer:
EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.