steel light
Setup:
• OpenClaw latest (just updated via npm update -g openclaw)
• macOS, Apple Silicon
• Channel: Telegram (direct chat)
• Model: Claude Opus
• Gateway: running fine, openclaw status shows all green
The issue:
My agent can send messages on Telegram perfectly, but every tool call (exec, apply_patch, ls, read, session_status, sessions_list — all of them) returns no result to the model. The calls seem to execute (I can sometimes see output on my end as the user), but the model never receives the result back. It just carries on as if no result was returned.
This means the agent can't read or write files, can't update memory, can't use any tools — essentially can't learn or persist anything.
What I've tried:
• Gateway restart (openclaw gateway restart)
• npm update to latest
• Multiple different tool calls (exec, apply_patch, ls, sessions_list, session_status)
• None return results
Terminal sessions seem unaffected — the issue appears to be Telegram-specific. The agent suggested it might be a session created before tools were properly configured, but there's no sessions reset or sessions delete command to test that theory.
Expected: Tool results feed back into model context
Actual: Tool calls fire but model receives nothing back
Any ideas? Is this a known Telegram integration issue?
real nimbus