Prototype an AI shopping agent that can complete a checkout on a user's behalf using a standard protocol.
Study a reference implementation of secure agent to business payment flows across Python, Go, and Android.
Integrate a business's checkout system with AI agents using AP2's declared capabilities model.
| google-agentic-commerce/ap2 | openmoss/moss-tts-nano | universal-commerce-protocol/ucp | |
|---|---|---|---|
| Stars | 3,001 | 3,032 | 2,961 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | easy |
| Complexity | 4/5 | 3/5 | 3/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires a Google API key or Vertex AI setup, plus Python 3.11+ and the uv package manager to run samples.
AP2, short for Agent Payments Protocol, is a specification and set of code samples for letting AI agents make purchases on a person's behalf in a secure, standardized way. Rather than every company building its own custom integration for agent driven shopping, AP2 defines a common protocol that platforms, businesses, payment providers, and credential providers can all speak, so agents can discover what a business supports and complete a checkout with or without a human present at that exact moment. The repository is organized into a docs folder holding the specification, flow diagrams, and FAQ, and a code folder holding the actual implementation. Inside code, there is an AP2 software development kit written in Python with Pydantic based data models and JSON schemas, a set of sample scenarios written in Python, Go, and for Android, and a demo web client built with Vite and React. Each sample scenario comes with its own README and a run script that installs dependencies and starts the relevant agents automatically. The samples are built using Google's Agent Development Kit and a Gemini model, though the README is clear that AP2 itself does not require either of these: developers are free to build their own agents with different tools. To try a scenario, you set up authentication using either a Google API key or Vertex AI, then run a scenario's script, for example the human present card payment flow, and navigate to the shopping agent's URL to interact with it. For developers who only want the protocol's data types rather than the full demo, the AP2 Python package can be installed directly from GitHub while a formal PyPI release is still pending. The project is released under the Apache 2.0 license and is maintained as an open, collaborative specification rather than a single company's closed product.
An open specification and code samples letting AI shopping agents complete secure, standardized payments on a user's behalf.
Mainly Python. The stack also includes Python, Go, React.
Apache 2.0: free to use, modify, and distribute, including commercially, as long as you keep the license and copyright notice.
Setup difficulty is rated moderate, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.