Alibaba Cloud Upgrades Token Plan Personal Edition With 12 Agent Harness Tools at Unchanged Prices

All of the tools are exposed through the MCP standard protocol, so developers can connect them directly to their own agent applications instead of buying each capability separately, according to the report.

The personal edition is aimed at individual developers and small and medium-sized teams. Its Lite tier costs 39 yuan a month and provides 2,500 Credits every seven days. Standard costs 139 yuan a month for 10,000 Credits every seven days, while Pro costs 499 yuan a month for 40,000 Credits every seven days.

With the upgrade, Standard and Pro subscribers receive monthly Harness allowances for seven MCP tools: enhanced web search, vertical search, webpage parsing, code interpreter, speech synthesis, speech recognition and image generation. The company said these cover common agent tasks such as retrieving material, processing webpages, analysing data and generating multimodal content.

Standard provides 100 calls of enhanced web search, 100 vertical searches, 1,000 webpage parses, 100 code interpreter runs, 10,000 characters of speech synthesis, five hours of speech recognition and 20 generated images each month. Pro provides 500 calls, 500 searches, 5,000 parses, 500 interpreter runs, 10,000 characters, five hours and 50 images on the same monthly basis.

Once the allowances included in a plan are used up, further calls are charged at 88 percent of the catalogue price, with the cost deducted from the user's Alibaba Cloud account.

Beyond the monthly benefits, Token Plan also added support for five services billed through post-payment: video generation, Managed Agent hosting, RAG knowledge base, Memory and Sandbox. The company said an agent task usually goes beyond model inference and also requires search, webpage parsing, speech processing, knowledge augmentation and code execution, while the memory an agent builds up can improve its results over time.

The upgrade turns the personal edition from a model quota subscription into a combination of model and agent services, placing model calls, code development, search, webpage parsing, image generation, speech processing and code execution on one plan and one bill. Capabilities such as Managed Agent, the RAG knowledge base and Memory are intended to carry a developer's agent from demo development to a production launch.

Taking a research task as an example, the company said a developer can collect material through search and webpage parsing, form structured conclusions, run data analysis in the code interpreter and produce supporting assets with image generation and speech synthesis, all within a single Standard or Pro plan. Token Plan is intended to lower the cost for individual developers to build a usable production-grade agent.

The new benefits took effect on September 11. Existing subscribers receive the added tools and a one-time seven-day quota reset, and users can subscribe through the Qwen AI platform. Token Plan can be called from Qwen desktop, Qoder, Qwen Code, Codex, OpenClaw and other agent applications, allowing developers to reuse the subscribed model across different tools.