Mcp Memory Multiuser

MCP Memory Query - multi-user (shared account)

This setup has two memory servers, configured because one agent account is shared by several people:

| Server | Tools | What it holds | |---|---|---| | memory-shared | mcp__memory-shared__* | Team/family-wide memory everyone sees | | memory-private | mcp__memory-private__* | This person's private memory only |

Hard isolation: the two servers point at separate databases with separate tokens. This person physically cannot read anyone else's private store, and others cannot read theirs. Privacy is enforced by the database boundary, not by tags. The owner: / scope: tags only describe a memory's intended audience.

If you only see one set of mcp__memory__* tools (no -shared / -private suffix), this is a single-user install - use the mcp-memory-query skill instead.

---

Reading: query BOTH stores, then merge

For any recall ("what do you remember about X", "find memories about X", "check your memory"), search both servers and combine the results:

1. mcp__memory-private__memory_search with the user's topic.
2. mcp__memory-shared__memory_search with the same topic.
3. Merge, de-duplicate, and present together. Label each hit with its origin so

the user knows what is private vs shared:

• Private hit -> tag it (private).
• Shared hit -> tag it (shared - visible to the whole team/family).

Use the same call conventions as the single-user skill:

• query: the topic as a short noun phrase.
• limit: 10 each by default; raise to 20-30 for a broad sweep.
• quality_boost: 0.3 when precision matters.
• max_response_chars: 30000 when result sets may be large.

For mistake notes / known pitfalls before a task, check mcp__memory-private__mistake_note_search first, then mcp__memory-shared__mistake_note_search.

Presentation:

Found N memories about "[topic]":

Private (only you):
1. [fact] - Tags / Type / Stored
Shared (whole team/family):
1. [fact] - Tags / Type / Stored

[If nothing found]: No memories stored about "[topic]" in your private or the
shared store. Want me to save one? (I'll keep it private unless you say it's for
everyone.)

Eventual consistency caveat (hybrid Cloudflare backend): a memory written on another device appears after the next background sync (seconds to minutes). If a lookup returns nothing the user expected, say it may not have synced yet rather than asserting it was never stored.

---

Writing: private by default, shared only on a clear signal

When the user says "remember this", "store this", "save this to memory", decide which store to write to. Default to private. This protects the user from leaking personal notes into a store the whole team/family can read.

Write to memory-shared only when the content is clearly meant for everyone. Signals:

• The user says so: "share this with the team", "everyone should know",

"save this to the family/team memory", "this is a shared convention".

• It is plainly a team/family-wide fact (a shared standard, a household rule, a

company-wide convention, a shared resource URL).

Write to memory-private (default) when:

• The user says "remember this for me", "my note", "just for me", or says

nothing about audience.

• The content is personal, role-specific, or about this person's own line of

work.

If it is ambiguous and might be sensitive, ask one short question before writing to shared: "Save this to your private memory, or share it with the whole team?" Never silently publish a personal note to the shared store - that is a one-way information leak.

Store call:

{
  "content": "",
  "metadata": { "tags": "tag1,tag2", "type": "note|decision|convention|reference" }
}

Use mcp__memory-private__memory_store or mcp__memory-shared__memory_store accordingly. Confirm back which store you used:

• Private: "Stored privately - only you can see this."
• Shared: "Stored in the shared memory - visible to the whole team/family."

To move a memory from private to shared later, store it in the shared store and (optionally) delete the private copy; there is no cross-store move tool.

---

Harvesting session learnings

If the user says "harvest memories" / "extract learnings from this session", run mcp__memory-private__memory_harvest with dry_run: true first and preview. Session learnings are personal by default, so commit them to the private store unless the user asks to share. (The post-session distillation hook already writes to the private store automatically.)

---

Edge cases

• Only one store responds: report which store is unavailable; still return

results from the one that worked.

• Nothing found: say so for both stores; offer to store (private by default).
• Too many results: ask the user to narrow by tag, date, or type.
• User asks "what does the team know about X": query only memory-shared.
• User asks "what do I have on X": query only memory-private.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: internetyev
• Source: internetyev/cross-agent-memory-kit
• License: MIT

Install and usage instructions live in the source repository linked above.