Vibeleaderboard
← All Intel
Intel / article

How we built Claude Code auto mode: a safer way to skip permissions

Source
anthropic.com
Author
Anthropic Engineering
Date

Context

Anthropic designed Claude Code auto mode to automate permission decisions while checking whether actions exceed the user's authorization. Its architecture separates two jobs: a server-side probe warns Claude about suspicious instructions in material it reads, while a classifier can block proposed tool calls. Reads and file edits inside the project skip that classifier, which reviews riskier actions such as shell commands. The engineering account explains why detecting hostile input and stopping an unauthorized action require different controls.

Those checks remain fallible. Anthropic reported that the complete pipeline missed 17% of 52 real examples of agents taking actions beyond what users intended. Auto mode therefore reduces, but does not remove, the need for users to judge what Claude should be allowed to do.

Terms in this piece · Glossary
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
  • context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
  • token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
  • chain-of-thought — Having a model write out intermediate reasoning steps before its answer, which markedly improves performance on hard problems.
Key quotes

“Claude Code users approve 93% of permission prompts. We built classifiers to automate some decisions, increasing safety while reducing approval fatigue.”

“Sandboxing is safe but high-maintenance: each new capability needs configuring, and anything requiring network or host access breaks isolation.”

“Auto mode uses two layers of defense: one for what Claude reads, one for what Claude does.”

Read the source www.anthropic.com
More from Anthropic Engineering
Recommended reads
Comments

Checking sign-in…

Loading comments…