I’m using OpenClaw via Telegram DM. When the assistant performs backend actions like editing files / reading memory / checking weather, some internal “working” narration appears in the Telegram chat UI
Expected: internal tool progress/narration should stay hidden unless explicitly user-facing.
Actual: backend/tool commentary is sometimes delivered as visible Telegram messages.
Setup:
OpenClaw 2026.4.24
Channel: Telegram direct chat
Model: openai-codex/gpt-5.5 / previously gpt-5.4
Runtime: OpenClaw Pi Default
Happy to provide screenshots/log snippets if helpful.