2026-10-22
Praxis is a platform for businesses to serve their own customers: Agents for every business.
Learn about Deepractice, how businesses and developers use the platform, our open-source and academic record, channels and recognition, our founder and core team, and how to reach us.
Helping businesses serve their own customers with agents
One platform, two ways to use it
A workspace for businesses and developers on the platform
Infrastructure for developers and enterprises to bring agents into their business
Open source puts methods to use; research puts them to the test
Model capacity sourced directly from first-tier upstreams
HK Tech 300 and industry recognition
Sean Jiang · Founder & CEO
Operations, product, and academic advisory
Phone, email, and website
Praxis is a platform for businesses to serve their own customers: Agents for every business.
We help businesses bring agents into their operations and serve their own customers. Customer relationships, data, and brands belong to the business.
Praxis makes three commitments: quick to integrate, affordable to use, and dependable. We help businesses connect agents to their websites and operations, pay for what they use, and sustain services for real customers.
We work by three principles: unity of knowledge and action (知行合一), governing by explicit rules (垂法而治), and compounding from small accumulations (积微成著).
Praxis is a platform for businesses to serve their own customers: Agents for every business.
One platform, two ways to use it.
Praxis Cloud provides infrastructure for developers and enterprises to bring agents into their business and serve customers on their own websites and in their operations. Models, search, documents, sandboxes, collaboration, and workflows are available on demand, without building the infrastructure yourself.
Praxis App is a workspace for businesses and developers to organize agent tasks and review progress and results on the platform, available on the web, desktop, and mobile.
Praxis is a platform for businesses to serve their own customers: Agents for every business.
Praxis App is a workspace for businesses and developers to organize agent tasks and review progress and results on the platform.
Praxis App is available on the web, desktop, and mobile. The agents you create each carry their own identity, role, and memory. They take on tasks, keep a record of their process, and deliver results.

Praxis App · agent collaboration workspace
Praxis is a platform for businesses to serve their own customers: Agents for every business.
Praxis Cloud helps developers and enterprises bring agents into their own websites and operations to serve their own customers.
Model routing, sandboxes, documents, agents, and more — used on demand with one account and one bill; usage, latency, and cost stay visible and controllable in the console, with nothing to build yourself.

Praxis Cloud · model routing console
Deepractice started from open source.
We maintain 43 public repositories on GitHub (as of August 2026). PromptX is the earliest and the best known — an AI agent context platform, open-sourced under the MIT license, with 3,687 stars (as of August 2026), continuously distributed through npm and Docker. In December 2025 it passed AtomGit's official review and received the G-Star program graduation certification (No.0478).
Beyond open source, we bring the methods of our practice back to academia for examination. A paper with the City University of Hong Kong, "PromptX: A Cognitive Agent Platform with Long-term Memory", was accepted by The Web Conference 2026 Demo Track.
For us these are two sides of the same thing: open source lets methods be genuinely used, and academia lets methods be rigorously examined.

Praxis Cloud's model capacity is purchased directly from Volcano Engine and Ant Group.
For an infrastructure company, who the upstream is is not a detail. If we provide model capacity to developers and enterprises, we have to guarantee that the supply is legitimate, traceable, and sustainable — not quotas pieced together from intermediaries of unknown origin.
Choosing first-tier vendors as direct upstreams is part of the reliability commitment: behind every model call made on Praxis Cloud stands a clear supplier with service guarantees.
In December 2025, Deepractice was selected into the City University of Hong Kong's HK Tech 300 seed fund program, receiving startup funding and incubation support.
In August of the same year, we presented our technical work "An AI Agent Memory System Based on Cognitive Psychology" at the official Cursor Nanjing Meetup, receiving official recognition and support from the Data Bureau of Jianye District, Nanjing.

HK Tech 300 · City University of Hong Kong

Founder & CEO
Software engineer and agent engineering specialist.
Formerly a testing, operations, and development engineer at WeBank, where he accumulated complete experience in large-scale internet engineering across development, testing, and operations. He then moved into internet solutions entrepreneurship, becoming deeplearning.ai's China partner, and later founded an internet travel business that reached a million-yuan annual profit with a team of 50–100.
After entering the AI field he became an active open-source contributor — author of popular GitHub AI open-source projects and founder of technology communities.
"Deepractice's philosophy is to make AI a tool that raises productivity, not a castle in the air."
— "Inspired by DeepSeek, he built new industrial productivity with 'Deep Practice'", Xinhua Daily · Jiaohu Point, reporter Shao Jia · Read the original

Lin Binya
Head of Operations
PhD in Literature from Tohoku University, Japan; MA in Literature from National Chengchi University, Taiwan; experienced in product operations, user growth, and ecosystem partnerships.

Qiu Yue (YoYo)
Head of Product
Former AI product manager at Yuanfudao Group and AI education content creator; double bachelor's in physics and psychology, master's in education from the University of Helsinki.

Dr. Wang Maolin
Academic Advisor
PhD from the City University of Hong Kong, author of 20+ CCF-A/B papers (AAAI/KDD among others), KDD 2025 best paper nominee; his EI-BERT model is deployed at Alipay serving hundreds of millions of users.