Hub/AI Security/Tool-call approval prompts
Which AI tools have tool-call approval prompts?
Asks the user before running tools, shell commands or MCP calls by default
9 Yes1 Partial
Verified Yes: Claude Code, Codex, Cursor, GitHub Copilot (VS Code), Windsurf / Devin Desktop, Gemini CLI, Kiro, Zed, Cline.
| Tool | Status | Notes | Source |
|---|---|---|---|
| Claude CodeAnthropic | Yes | Read-only by default; asks before Bash, file edits and MCP tools | Source |
| ClineCline | Yes | Asks before actions unless Auto-approve / YOLO is enabled | Source |
| CodexOpenAI | Yes | Approval policy (e.g. on-request); the untrusted policy was retired | Source |
| CursorAnysphere | Yes | Run modes; Auto-review runs allowlisted calls, sandboxes shell commands, sends the rest to a classifier | Source |
| Gemini CLIGoogle | Yes | Confirmations on by default; per-server trust:true bypasses them | Source |
| GitHub Copilot (VS Code)GitHub / Microsoft | Yes | Tool and terminal approvals; auto-approve and Autopilot are opt-in | Source |
| KiroAWS | Yes | MCP tools need approval unless listed in autoApprove | Source |
| Windsurf / Devin DesktopCognition | Yes | Prompts before any MCP tool call by default | Source |
| ZedZed Industries | Yes | Tool actions confirm by default (always-allow defaults to off) | Source |
| JetBrains AIJetBrains | Partial | IDE MCP server confirms commands from external clients unless brave mode is on | Source |
| ChatGPT (apps & connectors)OpenAI | Unknown | help.openai.com blocks automated fetches; not verified | — |
| Claude apps (claude.ai / Desktop)Anthropic | Unknown | Not found in official docs checked 2026-11-04 | — |
Other AI Security features
Weekly, free
What changed in AI integrations
One short email when MCP clients, plugin ecosystems, IDE integrations, AI plan prices or security advisories change. Every item is sourced.