langchain-channel3

Channel3 product search tools for LangChain. Give your agent search across 100M+ products from thousands of retailers, with every retailer's offer and live price.

Installation

pip install -U langchain-channel3

Set CHANNEL3_API_KEY in your environment. Create a free API key: every account gets 1,000 free credits a month. A search uses one credit; product details by ID are free, and each retailer URL lookup uses one credit.

Usage

from langchain.agents import create_agent
from langchain_channel3 import channel3_tools

agent = create_agent(
    model="anthropic:claude-sonnet-5-5",
    tools=channel3_tools(),
)

result = agent.invoke(
    {
        "messages": [
            {
                "role": "user",
                "content": "Find me a leather golf glove under $40 and tell me which retailer has it cheapest.",
            }
        ]
    }
)
print(result["messages"][-1].content)

Tools

Channel3SearchProducts (search_products)

Searches for one product type and returns in-stock products, each with up to two images and every retailer's offer.

• query: one product type and its constraints in natural language, such as "red leather jacket under $200".
• image_url: a public image URL for visual search. Combine it with query for text plus image.

At least one is required.

Channel3GetProducts (get_products)

Gets full product data (description, attributes, variants, images and offers) for up to 40 products in parallel.

• product_ids: product IDs from search_products, or retailer product URLs.

Inputs that can't be resolved are listed in unresolved. The call fails only when none resolve.

Both tools use response_format="content_and_artifact". For search_products the model sees one line per product with title, brand, best price and ID; for get_products it sees JSON. The full structured result is on the ToolMessage as artifact, so your UI can render product cards from it.

Offer URLs are Channel3 buy links, so purchases earn affiliate commission for the API key's account.

Options

from langchain_channel3 import Channel3Toolkit

tools = Channel3Toolkit(api_key="...", conversation_id=chat_id, limit=12).get_tools()

The tool classes take the same options individually, for example Channel3SearchProducts(limit=12).

channel3_tools(**options) is shorthand for Channel3Toolkit(**options).get_tools().

Errors

API errors go back to the model as the tool result, for example "Channel3 rejected the API key" or "The Channel3 account is out of credits", so the agent can explain the problem instead of crashing the run. Pass handle_tool_error=False to a tool to raise them as ToolException instead.

Links

• Channel3 docs
• TypeScript version for LangChain.js, the Vercel AI SDK and Mastra
• Channel3 Python SDK for the rest of the API

License

MIT