Summary
On memory-lancedb-pro@1.1.0-beta.9, memory_update still fails for an agent:main memory that was just created and is clearly accessible.
Observed behavior:
memory_store succeeds
memory_list can see the memory
memory_recall can find the memory
memory_forget can delete the same memory ID
- but
memory_update returns: Memory <id>... not found or access denied.
So this does not look like a full plugin failure. It looks like the update path is still broken or inconsistent, likely around ID/scope/context resolution.
Environment
- Plugin:
memory-lancedb-pro@1.1.0-beta.9
- OpenClaw:
2026.3.13
- OS: Windows 11
- Scope used:
agent:main
- Current runtime confirms plugin is loaded and tools are exposed.
Reproduction
- Store a memory:
memory_store(text="MEMORY_UPDATE_RETEST_20260320_2052", importance=0.53, category="fact", scope="agent:main")
- Confirm it exists:
memory_list(limit=15, scope="agent:main", category="fact")
- returned ID:
3fd0304d-3e9c-49ef-8873-6ffb16ebdb24
- Try updating the exact same ID:
memory_update(memoryId="3fd0304d-3e9c-49ef-8873-6ffb16ebdb24", text="MEMORY_UPDATE_RETEST_20260320_2052_UPDATED", importance=0.54, category="fact", scope="agent:main")
- Actual result:
Memory 3fd0304d... not found or access denied.
- Then immediately try forgetting the same ID:
memory_forget(memoryId="3fd0304d-3e9c-49ef-8873-6ffb16ebdb24", scope="agent:main")
- Forget succeeds.
Why this seems important
This suggests the problem is specifically in memory_update, not in general memory visibility/access:
- same memory can be listed
- same memory can be recalled
- same memory can be forgotten
- but same memory cannot be updated
Possibly related upstream work
I noticed these existing items seem related:
However, this issue is being observed still in beta.9 runtime behavior during live testing on Windows with agent:main scope.
Expected behavior
If a memory ID is valid enough to be listed/recalled/forgotten within the same scope/context, memory_update should also be able to update it.
Actual behavior
memory_update reports not found or access denied for a memory that is otherwise clearly accessible.
Summary
On
memory-lancedb-pro@1.1.0-beta.9,memory_updatestill fails for anagent:mainmemory that was just created and is clearly accessible.Observed behavior:
memory_storesucceedsmemory_listcan see the memorymemory_recallcan find the memorymemory_forgetcan delete the same memory IDmemory_updatereturns:Memory <id>... not found or access denied.So this does not look like a full plugin failure. It looks like the update path is still broken or inconsistent, likely around ID/scope/context resolution.
Environment
memory-lancedb-pro@1.1.0-beta.92026.3.13agent:mainReproduction
memory_store(text="MEMORY_UPDATE_RETEST_20260320_2052", importance=0.53, category="fact", scope="agent:main")memory_list(limit=15, scope="agent:main", category="fact")3fd0304d-3e9c-49ef-8873-6ffb16ebdb24memory_update(memoryId="3fd0304d-3e9c-49ef-8873-6ffb16ebdb24", text="MEMORY_UPDATE_RETEST_20260320_2052_UPDATED", importance=0.54, category="fact", scope="agent:main")Memory 3fd0304d... not found or access denied.memory_forget(memoryId="3fd0304d-3e9c-49ef-8873-6ffb16ebdb24", scope="agent:main")Why this seems important
This suggests the problem is specifically in
memory_update, not in general memory visibility/access:Possibly related upstream work
I noticed these existing items seem related:
memory_update rejects agent-scoped memories as outside accessible scopesmemory_forget fails with 'outside accessible scopes' even when agent should have accessfix: use resolveToolContext for scope resolution in forget/update/stats/list toolsfix: use resolveToolContext in memory_forget and memory_updateHowever, this issue is being observed still in beta.9 runtime behavior during live testing on Windows with
agent:mainscope.Expected behavior
If a memory ID is valid enough to be listed/recalled/forgotten within the same scope/context,
memory_updateshould also be able to update it.Actual behavior
memory_updatereportsnot found or access deniedfor a memory that is otherwise clearly accessible.