Claude Desktop: allow auto-approving Analysis tool execution

Feature Request

Problem

When Claude Desktop uses the Analysis tool (built-in Python code execution) repeatedly, the user is prompted with "Allow once / Deny" for every single execution. In workflows where Claude runs many similar Python scripts back-to-back (e.g., iterating on API calls, data processing), this becomes extremely tedious.

The permission dialog shows the full script source, but realistically nobody reads each one when they're all structurally similar commands. It's just friction without added security value in these cases.

Current Behavior

• Every Analysis tool invocation shows a permission prompt
• User must click "Allow once" or press ⌘+Enter each time
• No option to auto-approve for the conversation or session
• No alwaysAllow equivalent like MCP servers have

Requested Behavior

Add an option to reduce or eliminate repeated prompts, for example:

• "Allow for this conversation" — auto-approve all Analysis tool runs in the current chat
• "Always allow" — persistent setting to never prompt for Analysis tool
• "Allow N times" — batch-approve upcoming executions

This is similar to how MCP servers support alwaysAllow in claude_desktop_config.json — the Analysis tool should have a comparable mechanism.

Use Case

User is debugging Home Assistant websocket API calls. Claude runs 10+ similar Python scripts sequentially, each making a websocket connection and testing different parameters. Every script requires manual approval despite being nearly identical.

🤖 Generated with Claude Code