Claude Developers Launch Code Mode, an Experimental Way for Claude Code to Use Tools
Check out Code Mode, a new, experimental way for Claude Code to work with tools. Claude writes scripts that chain your tools together in sophisticated ways, and it can answer you while long commands are still running. Opt in with CLAUDE_CODE_EXPERIMENTAL_CODE_MODE=1
Context
Anthropic is letting Claude Code write scripts that chain tools together and keep talking while long commands are still running, but it is still an opt in experiment.
Claude Developers posted that Code Mode is a new, experimental way for Claude Code to work with tools. The post says Claude writes scripts that chain tools together in sophisticated ways, and that it can answer while long commands are still running. Users opt in by setting CLAUDE_CODE_EXPERIMENTAL_CODE_MODE=1. A documentation link is in the post.
Deep dive
The official Claude Developers account announced Code Mode as an experimental feature for Claude Code, Anthropic's coding agent. According to the post, Claude writes scripts that chain a user's tools together, and it can keep answering while long commands are still running. That is a change from a pattern in which an agent calls one tool at a time and waits, at least as the post describes it.
The feature is not on by default. The post tells users to opt in with the environment variable CLAUDE_CODE_EXPERIMENTAL_CODE_MODE=1, and it points to documentation at code.claude.com. Calling it experimental is Anthropic's own label, so this should not be read as a finished default for Claude Code.
The post does not give benchmarks, pricing changes, or a list of which tools can be chained. It also does not say when, or whether, Code Mode will leave the experimental flag. Those details, if they exist, are in the linked docs or in later posts, not in the announcement itself.
This is an official product note from the Claude Developers account, not a leak or a third party writeup. It is a tooling change for people already in Claude Code, not a new frontier model.
What to expect next
Usage under the experimental flag will show whether scripted tool chaining is stable enough to become a default. Watch the documentation for supported tools, failure cases, and any later decision to remove the flag. If Anthropic posts examples of long running commands that stay interactive, those will be the practical test. Rival coding agents will be compared on the same idea of chaining tools in a script rather than in a single step loop.
