memory-status

Report where memory actually lives and how to connect, so the human never inspects an empty root while the agent wrote elsewhere.

1. Print the resolved data root (the console applies --root > LM_DATA_ROOT > ~/.lean_memory):

   !echo "Resolved root: ${LM_DATA_ROOT:-$HOME/.lean_memory}"

   Then enumerate namespace .db files under that root (skipping _*.db):

   !ls -1 "${LM_DATA_ROOT:-$HOME/.lean_memory}"/*.db 2>/dev/null | grep -v '/_' || echo "(no namespaces yet)"

2. ./lm_data mismatch warning. The core engine's own default root is ./lm_data, not ~/.lean_memory. If ./lm_data exists in the current project but is not the served root, the human would silently inspect an empty root. Warn when it exists:

   !test -d ./lm_data && echo "WARNING: ./lm_data exists here. Your agent may have written memories to ./lm_data (the engine's default root), not ${LM_DATA_ROOT:-$HOME/.lean_memory}. Run '/memory:ui' with --root ./lm_data to inspect it." || echo "No ./lm_data in this directory."

3. Connect snippets.

   • Local observing MCP (this plugin already wires it via .mcp.json): uvx lean-memory-console mcp
   • Open the console: /memory:ui
   • Docker HTTP (after /memory:server-up):

     claude mcp add --transport http lean-memory http://127.0.0.1:8377/mcp \
       --header "Authorization: Bearer $LM_API_KEY"
     

Guidance: use one namespace per project/session — cross-process writers on a single namespace serialize via retry, not a lock manager.