Troubleshooting OpenClaw Memory: Forgetfulness, Slow Context, and Storage Bloat

Start here: Run openclaw doctor and openclaw status before deep edits. See quick diagnostics.

Symptom → diagnosis flow

Symptom	Likely cause	First fix
Agent "forgot" yesterday	History limit too small; wrong agent memory scope	Raise `historyLimit`; check per-agent memory
Slow replies	Oversized context sent every turn	Enable summarization; trim MEMORY.md; see patterns
Disk growing fast	Unbounded session logs	Set retention; archive old sessions; backup then prune
Conflicting facts	Multiple channels writing same USER.md	Split agents or namespace memory files

Diagnostic commands

openclaw doctor
openclaw status
openclaw config validate
openclaw logs --tail 100
# Check memory directory size (Linux/macOS)
du -sh ~/.openclaw/memory 2>/dev/null || du -sh %USERPROFILE%\\.openclaw\\memory

Fix checklist

Confirm you are messaging the same agent ID across channels.
Verify persistPath is writable (common Docker volume issue — Docker install).
After upgrades, re-run migration steps.
For model-side limits, pick a higher-context stack (LLM guide).

Still stuck? Cross-reference error codes and base memory guide.