Overview
Claude Commerce Agents is Anthropic's reference implementation of two commerce agents built on Claude. The shopping agent is what a business embeds in its app for customers: it searches and compares products, plans, fills the cart, answers order and policy questions and remembers what a customer tells it. The merchant agent is what staff use to run the back office: it explains performance, maintains listings, acts on inventory and order alerts, prices and promotes, and drafts campaigns. Each agent is defined once — prompt, skills, tool contracts and gates — and runs on the Messages API, the Claude Agent SDK and Managed Agents.
The repository ships seven pip packages (a shared commerce-common plus core, runtime and SDK layers per agent) and four runnable verticals — retail, travel, telecom and entertainment ticketing — each with a storefront and a merchant portal over the same libraries. Everything runs against fictional stores: nothing places an order, charges a card or changes a live listing. Checkout renders the cart for the host application to complete, and every merchant write is staged until a person approves it.
Safety is enforced inside the tool calls rather than in the prompt: fencing of third-party text, provenance gates so only IDs returned by a catalogue tool can enter a cart, caps, memory validation and the merchant approval gate hold on all three execution paths. A Claude Code plugin scaffolds a new agent on these packages against your own systems or reviews one you already have. Anthropic labels the project a reference implementation under Apache-2.0 that is not maintained and does not accept contributions — a starting point you own.
What it does
- Two reference agents: a customer-facing shopping agent and a back-office merchant agent, each defined once and run on the Messages API, the Agent SDK and Managed Agents
- Four runnable verticals (retail, travel, telecom, entertainment) with a storefront and a merchant portal each
- Guardrails in code: text fencing, provenance gates on cart and listing IDs, caps, memory validation and a human-approval gate on every merchant write
- Checkout hands off to the host — the agent never sees payment details or the checkout URL
- A Claude Code plugin (commerce-builder) that scaffolds, extends, evaluates and reviews commerce agents
- Backend interfaces (StorefrontBackend, MerchantBackend) you implement over your own catalogue, cart, order, analytics and pricing systems; deployment docs cover Vertex AI, Bedrock and Microsoft Foundry
Getting started
You need Python 3.11+ and Node 22. The README's quick start clones the repo, installs the packages and runs a vertical; the commands below are taken from it.
Clone and install the packages
The pinned requirements install the seven packages from their directories.
git clone https://github.com/anthropics/commerce-agents.git && cd commerce-agents
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txtAdd your API key and install the web apps
Copy the example environment file and set ANTHROPIC_API_KEY; the eight web apps share one npm workspace.
cp .env.example .env
(cd examples && npm ci)Run the retail demo
Starts the API on port 8000 and the storefront on port 3000. Pass --merchant for the portal instead, --all for both; the other verticals are travel, telecom and entertainment.
python scripts/run_demo.py retailScaffold your own agent with the Claude Code plugin
With the repo cloned, the plugin reads it as the reference and builds a project against your stack.
claude plugin marketplace add anthropics/commerce-agents
claude plugin install commerce-builder@claude-commerce-agents
claude
/scaffold-commerce-agent a shopping assistant for our storeCommands and code are distilled from the project's own documentation — always check the official repo for the latest.
When to use it
- Start a retail or marketplace shopping assistant from a build whose safety rules are already enforced in code
- Give merchant staff an agent that drafts pricing, promotion and restock changes for human approval instead of applying them
- Study how Anthropic wires the same agent onto the Messages API, the Agent SDK and Managed Agents
- Review an existing commerce agent against the reference with the plugin's /review-commerce-agent command
How Claude Commerce Agents compares
Claude Commerce Agents alongside other open-source agent frameworks & builders tools AI/TLDR tracks, ranked by GitHub stars.
| Tool | Stars | What it does |
|---|---|---|
| DeepSeek Harness | ★ 213k | DeepSeek AI's open-source agent harness (dsh), built on Cordis, where models, tools, skills, sessions, sandboxes, storage and the UI are all plugins composed through profiles. |
| AutoGPT | ★ 187k | One of the earliest autonomous agent projects, now a platform for building and running agents from reusable blocks and workflows. |
| DeerFlow | ★ 81.4k | ByteDance's open-source super agent harness built on LangGraph: skills, sub-agents, sandboxes, a filesystem and long-term memory for long-horizon research, coding and content tasks. |
| nanobot | ★ 47.7k | Lightweight self-hosted personal AI agent framework in Python, with a WebUI, terminal and chat-app channels, tools, long-term memory, MCP and scheduled automations. |
| Agno | ★ 42.1k | A fast Python framework (formerly Phidata) for building agents with memory, tools, and multimodal inputs, plus a runtime for deploying them in production. |
| LangGraph | ★ 41.1k | A library from the LangChain team for building stateful, graph-based agent workflows with explicit control over steps, memory, and human-in-the-loop checkpoints. |
| AgentGPT | ★ 36.3k | AgentGPT lets you name a custom AI, give it a goal, and watch it plan tasks, run them, and learn from the results, all from a web browser. |
| Claude Commerce Agents | ★ 2k | Anthropic's open reference build of a shopping agent and a merchant agent on Claude, with guardrails enforced in code |
