Memory banks
Banks are isolated memory stores. Each bank is a separate database file with its own memories, sessions, and statistics. Memories in one bank are invisible to another.
Why use multiple banks?
| Use case | Strategy |
|---|---|
| Per-project | One bank per codebase, switch via the bank parameter or --bank flag |
| Per-topic | One bank for cooking, another for finance, another for travel |
| Per-environment | One bank for dev, one for production data |
| Per-user | If you share a machine, separate banks per user |
| Per-purpose | Short-term context (auto-cleaned) vs. long-term knowledge (persistent) |
The default bank is always available — it’s created on first run. Use additional banks when you want hard isolation.
Naming
Bank names are free-form strings but follow these rules:
- Allowed: letters, digits,
-,_ - Avoid spaces (use
-or_) - The name
defaultis reserved - Case-sensitive
Good names: project-x, research-2026, meeting-notes, personal.
Creating banks
Banks are created automatically on first write to a new name. The server handles this transparently:
// MCP call to add_content_memory
{
"content": "Postgres over MongoDB for analytics",
"bank": "project-x" // <-- creates project-x if it doesn't exist
}
You can also pre-create with create_memory_bank (MCP) or mkdir + populate on disk.
Listing banks
# CLI
llm-mem list-banks
// MCP: list_memory_banks
{}
Output:
{
"banks": [
{ "name": "default", "memory_count": 1234, "size_bytes": 45678901 },
{ "name": "research", "memory_count": 567, "size_bytes": 12345678 }
],
"count": 2
}
Switching the active bank
MCP: pass bank: "<name>" on every call. There’s no global “active bank” concept at the protocol level.
CLI REPL: use <bank> switches the active bank for the session.
CLI single-command: --bank <name> per command.
Isolation guarantees
- Cross-bank searches are impossible (a tool can only see one bank at a time)
- Cross-bank relations don’t exist
- The abstraction pipeline operates per-bank (L0→L1 in bank A doesn’t see content from bank B)
- Backups and exports are per-bank
Soft grouping within a bank
For softer grouping without full isolation, use context tags and topics:
context— broad topic tags (e.g.["project-x", "auth", "jwt"])topics— specific subjects within the content
You can then filter by context/topic in queries:
// MCP: query_memory
{
"query": "JWT auth",
"context": ["project-x"]
}
Bank vs user_id
user_id is a per-memory filter within a bank. It’s optional and only useful when multiple users share a single bank.
In most cases:
- Single user per bank: omit
user_id - Multiple users per bank: set
user_idto differentiate
If two users share a bank and you don’t filter by user_id, each will see the other’s memories. Use a separate bank if you want hard isolation per user.
Lifecycle
| Operation | Tool | CLI |
|---|---|---|
| Create | create_memory_bank | (auto, or mkdir) |
| List | list_memory_banks | list-banks |
| Rename | rename_memory_bank | db rename |
| Backup | backup_bank | db export |
| Restore | restore_bank | db import |
| Delete | cleanup_resources (target: “banks”) | rm -rf the bank directory |