Claude Code introduces an experimental Code Mode: multiple tools can be used in one call, and long tasks can run in the background
BIBIBI
AT A GLANCE
According to a quick update, Claude Code requires manual activation of the experimental Code Mode to support script calls for multiple tools and long background tasks.
Article
Claude Code launches experimental Code Mode: multiple tools can be called at once, and long tasks can run in the background
Beating AI Flash News: Anthropic has added an experimental Code Mode to Claude Code, which went live in v2.1.323 released on November 12 and currently needs to be manually enabled.
Once enabled, Claude can first write JavaScript scripts, then use these scripts to call multiple tools continuously and in parallel. It can also process the data returned by the tools beforehand, passing only the necessary results to the model, thereby reducing model requests and Token consumption.
The new mode also supports long-running tasks executing in the background. Claude can execute commands while continuing to answer the user. Official documentation indicates that background scripts can run for up to 2 hours, but require automatic authorization modes such as Auto.
Code Mode was proposed by Cloudflare in September 2025, and Anthropic unveiled a similar solution in November of the same year. This year, open-source Agent frameworks such as DeepSeek Harness, Pi, and OpenCode 2.0 have successively added support.
Anthropic is only now integrating it into Claude Code; it remains in the experimental stage and requires manual activation.
According to a news report, Anthropic added an experimental Code Mode to Claude Code v2.1.323 released on November 12, and it must be enabled manually at present.
02
After it is enabled, Claude can write JavaScript scripts, call multiple tools sequentially or in parallel, and process the data returned by the tools first, delivering only the necessary results to the model.
03
The original text states that this method can reduce model requests and Token consumption.
04
According to official documents, the backend script can run for up to 2 hours, but an automatic authorization mode such as Auto is required.
05
It is said that Cloudflare introduced Code Mode in September 2025, and Anthropic revealed a similar solution in November of the same year.
06
It is said that open-source Agent frameworks such as DeepSeek Harness, Pi, and OpenCode 2.0 have been added to support this year.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
According to reports, Claude can first write a script that uses multiple tools in sequence or simultaneously, and then pass the filtered necessary results to the model. Long tasks can also be placed in the background, allowing Claude to continue answering users during execution. This feature is not a stable function that is enabled by default.
Why it matters to readers
This update provides an alternative way to call organizational tools, and reports highlight its ability to reduce model requests, lower Token consumption, and support long-backend tasks; the actual effectiveness still needs to be tested.
New users should note that it is still in the experimental stage and needs to be enabled manually; the support for background tasks lasting up to 2 hours also requires an automatic authorization mode.
Risks and unknowns
The evidence provided is a summary from a quick message; it does not include the full text of official documents or version release records, and the relevant details have not been independently verified.
The original text does not provide model request or Token consumption reduction test data.
The original text does not specify the stability of experimental features, the specific activation steps, or the permission boundaries for automatic authorization.
Related Developments
Loading event timeline…
Related concepts
Code Mode
This term is not in the glossary yet. Browse related concepts in the glossary.