How to add Claude Code-style PreToolUse/PostToolUse hooks to Qwen Code?

Question

I want to control what happens before and after each tool call in Qwen Code, like Claude Code's PreToolUse/PostToolUse hooks — for example to block destructive bash commands, force an MCP tool for files with secrets, or auto-format after edits. A GitHub issue requested this feature and proposed a concrete protocol. Has it been implemented, and what config does it use?

Originally asked on GitHub: https://github.com/QwenLM/qwen-code/issues/268

AgentArmorHQ team · Accepted Answer

**Status: closed as completed.** A maintainer confirmed in March 2026 that a basic hooks implementation was submitted in PR #1988, with more hooks to follow in #2033. The thread does not state which Qwen Code version includes it or whether the final protocol matches the proposal below.

The thread's proposal (modeled on Claude Code/Cline) is the concrete reference. Add a `hooks` block to `settings.json`:

```json
{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [{ "type": "command", "command": "~/.qwen/hooks/bash-guard.sh" }]
      }
    ],
    "PostToolUse": [
      {
        "matcher": "Write|Edit",
        "hooks": [{ "type": "command", "command": "~/.qwen/hooks/auto-format.sh" }]
      }
    ]
  }
}
```

Proposed protocol: the hook receives JSON on stdin with `{"tool_name": ..., "tool_input": ...}`; `matcher` supports pipe-separated tool names; PreToolUse exit 0 allows and exit 2 blocks, with stdout as the message to the agent; PostToolUse exit code is ignored.

Use cases from the thread: enforce a local MCP redaction server so native Read/Write/Edit never expose secrets, block risky Bash commands, auto-format after edits, and a Stop hook that checks whether tests ran. The commenter noted they recommended Cline over Qwen Code for sensitive-data work because of this gap.

Security note: the core motivation is structural enforcement so agents can't send raw secrets to cloud APIs or run destructive commands; treat hook scripts as sensitive code and review them before use.

AgentArmorHQ team · Answer

**What @ppiankov tried on GitHub** (Outcome unclear)

Proposed adopting the Claude Code/Cline hook protocol: settings.json hooks config with matcher/command, JSON on stdin, exit 0=allow / exit 2=block for PreToolUse, exit ignored for PostToolUse. Use cases: enforce MCP redaction tools for secret files, block dangerous bash, auto-format after edits, and a Stop hook for test verification.

Source: https://github.com/QwenLM/qwen-code/issues/268#issuecomment-3966262415

How to add Claude Code-style PreToolUse/PostToolUse hooks to Qwen Code?

2 answers

Add your answer or comment

How to add Claude Code-style PreToolUse/PostToolUse hooks to Qwen Code?

2 answers

Add your answer or comment

Related questions