Issue: Missing Agent Status Visibility in Webchat During Tool Execution

Summary

The webchat interface has no way to indicate the agent is actively working during tool calls, making it impossible for users to distinguish between "agent is working" and "agent is dead/hung."

Current Behavior

1. When the agent invokes tools (exec, read, file operations, etc.), the webchat UI completely hides the agent's response
2. No typing indicator, spinner, or status message persists across the tool execution
3. The UI appears frozen — users see nothing while tools are running
4. Any subsequent user messages queue invisibly during the silent turn
5. There is no cancel/abort mechanism during tool execution
6. The only recovery option is to refresh the page

What is Needed

1. Persistent Status Indicator

• A small badge, spinner, or text overlay that persists across ALL tool invocations in a turn
• Should say something like "agent is working" or "processing..."
• Must survive transitions between multiple tool calls (agent may call 3-5 tools in sequence)

2. Interim Progress Updates

• Allow the agent to send status messages during tool execution
• Similar to how Telegram/Discord show "agent is typing..." while actually working
• Example: "checking Ollama config..." → "updating OpenClaw..." → "restarting service..."

3. Abort/Cancel Button

• Visual button that immediately interrupts the current turn
• Currently messages queue invisibly — need an explicit abort mechanism

4. Better Connection Status

• The current WebSocket flashing indicator does not distinguish between:
  • Agent is actively working (tools executing)
  • Agent is dead/hung (connection lost)
  • Agent is waiting for user input
• Should show clear state transitions

Why This Matters

User Experience Impact

• Power users doing complex multi-step admin work (system configuration, agent management, debugging) cannot see progress
• The silence creates genuine uncertainty: is the agent working or broken?
• Users develop workarounds like refreshing pages, which is poor UX

Inconsistency Across Interfaces

• Telegram interface: shows "Claw is typing..." status during tool execution ✅
• Discord interface: shows "Claw is typing..." status during tool execution ✅
• Webchat interface: shows NOTHING during tool execution ✗

Blocked Use Cases

• Long-running operations (service restarts, config changes, model operations)
• Debugging sessions where users need to monitor progress
• Any admin work requiring trust that the agent is still alive
• Users who cannot easily switch to Telegram/Discord

Scope

• This is a frontend/webchat issue, not a gateway config limitation
• The webchat was built as a simple chat UI for Q&A, not for multi-step operations
• The underlying agent/tool infrastructure works fine — the display layer is missing

Priority: High

This blocks effective use of webchat for complex operations while Telegram/Discord work without issues. Power users are forced to either:

1. Use a different interface entirely
2. Develop manual workarounds (constant refreshing, external monitoring)
3. Lose confidence in the system

The Telegram/Discord interfaces already solved this problem with typing indicators. Webchat should have an equivalent mechanism.

Implementation Suggestions

Quick Fix

• Add a persistent "Agent is working" overlay/badge during all tool execution phases
• Show the current tool being executed (like a status line)

Medium Fix

• Allow the agent to send interim status messages during turns
• Add a visual cancel button during tool execution

Full Solution

• WebSockets with status events: agent:working, agent:completed, agent:error
• Frontend shows real-time progress updates
• Proper state machine for session lifecycle

Environment

• OpenClaw version: 2026.4.22
• Interface: webchat (default gateway UI)
• Operating system: Linux (Ubuntu 24.04, ARM64)
• Browser: Modern Chromium-based