MAX_THINKING_TOKENS

Updated October 13, 2026 by Alex Sorokoletov

MAX_THINKING_TOKENS is a Claude Code environment variable: Likely controlled via API settings now. Its release history is not tracked here. It is listed in the official reference.

Status	Present in v2.1.294
Category	Authentication
Official docs	Listed

How to set it

For one session, export it in the shell that starts Claude Code:

export MAX_THINKING_TOKENS=<value>
claude

To apply it to every session, add it to the env object of a settings file:

// ~/.claude/settings.json
{
  "env": {
    "MAX_THINKING_TOKENS": "<value>"
  }
}

Release notes that mention it

Related variables

CLAUDE_CODE_FILE_READ_MAX_OUTPUT_TOKENS: Max tokens for file read output.
CLAUDE_CODE_MAX_CONTEXT_TOKENS: Sets an explicit ceiling on the context window size used during a session.
CLAUDE_CODE_MAX_OUTPUT_TOKENS: Maximum output tokens; validated per-model.
MAX_MCP_OUTPUT_TOKENS: Max tokens for MCP tool output.
CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS: Sets the minimum conversation size, in tokens, before idle compaction runs.
CLAUDE_BG_SOCKET_TOKENS_PATH
CLAUDE_CODE_TOTAL_TOKENS_REMINDER
CLAUDE_CODE_TOTAL_TOKENS_REMINDER_AFTER_USER_TURN

Frequently asked questions

What does MAX_THINKING_TOKENS do?

Likely controlled via API settings now

Which Claude Code version has MAX_THINKING_TOKENS?

Its release history is not tracked here.

How do I set MAX_THINKING_TOKENS?

Export it before starting Claude Code (export MAX_THINKING_TOKENS=value), or add it to the "env" object in ~/.claude/settings.json to apply it to every session.

All variables, searchable: Claude Code environment variables.