Skip to main content
Best Practices

Integrate Harness tools

Extend AI coding tools with built-in Harness tools in supported Qwen models

Some Qwen models supported by Token Plan have built-in Harness tools that can extend AI coding tools with capabilities such as web search, code interpreter, and web scraping.
Applicable to Token Plan, not applicable to Coding Plan.

Tool overview

ToolDescription
Web searchRetrieves information from the internet and generates answers based on search results.
Code interpreterWrites and runs Python code in a sandbox environment for scenarios such as mathematical calculations and data analysis.
Web scrapingAccesses a specified URL and extracts its content to provide the model with the information it needs.
Reverse image searchSearches for visually similar images on the internet based on an input image. Suitable for scenarios such as finding similar products and visual content tracing.
Text-to-image searchSearches for relevant images on the internet based on text descriptions. Suitable for scenarios such as visual Q&A and image recommendation.

Supported models and tools

Personal Edition

ModelSupported tools
qwen3.8-max-previewWeb search, code interpreter, web scraping, reverse image search, text-to-image search
qwen3.7-maxWeb search, code interpreter, web scraping
qwen3.7-plusWeb search, code interpreter, web scraping, reverse image search, text-to-image search

Team Edition

ModelSupported tools
qwen3.8-max-previewWeb search, code interpreter, web scraping, reverse image search, text-to-image search
qwen3.7-maxWeb search, code interpreter, web scraping
qwen3.7-plusWeb search, code interpreter, web scraping, reverse image search, text-to-image search

Pricing

Harness tools are billed per successful invocation, and fees are deducted from your plan Credits.

Usage

Switch the model in your AI coding tool to one of the Qwen models listed above that supports Harness, and ask questions directly in the conversation. The model will automatically invoke the corresponding built-in tool based on the question, without any additional configuration.