memory_forget

Delete a memory node from the graph. Optionally cascade to delete connected edges.

Parameters

The ID of the node to delete.

If true, also delete all edges and hyperedges connected to this node. Default: false.

Target graph. Defaults to your account’s default graph.

Example request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "memory_forget",
    "arguments": {
      "node_id": "node_abc123",
      "cascade": true
    }
  }
}

Example response

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "id": "node_abc123",
    "status": "deleted",
    "deleted_at": "2026-03-15T14:30:00Z",
    "edges_deleted": 3,
    "hyperedges_affected": 1
  }
}

Response fields

Field	Type	Description
`id`	string	Deleted node’s ID
`status`	string	Always `"deleted"` on success
`deleted_at`	string	Deletion timestamp
`edges_deleted`	number	Count of edges deleted (when cascade=true)
`hyperedges_affected`	number	Count of hyperedges this node was removed from

Errors

Code	Description
`INVALID_PARAMETER`	Missing `node_id`
`NODE_NOT_FOUND`	Node doesn’t exist
`GRAPH_NOT_FOUND`	Specified `graph_id` doesn’t exist

Notes

Free operation — Deletions don’t count against your query limit.

Cascade behavior

cascade: false (default)

Node is deleted
Connected binary edges become “orphaned” (point to non-existent node)
Hyperedges have this node removed from their nodes array
Orphaned edges may cause issues — use cascade: true for clean deletion

cascade: true

Node is deleted
All binary edges connected to this node are deleted
Hyperedges have this node removed from their nodes array
If a hyperedge would have fewer than 3 nodes, it’s converted to a binary edge or deleted

Hyperedge handling

When a node is removed from a hyperedge:

Remaining nodes	Result
3+ nodes	Hyperedge continues with reduced membership
2 nodes	Hyperedge converts to binary edge
1 node	Hyperedge is deleted

Permanent deletion — There is no undo. For recoverable deletion, consider adding metadata like status: "archived" instead.

Batch deletion

To delete multiple nodes, call memory_forget for each:

for node_id in nodes_to_delete:
    client.call_tool("memory_forget", {
        "node_id": node_id,
        "cascade": True
    })

Soft delete pattern

Instead of deleting, mark nodes as archived:

# Soft delete
client.call_tool("memory_update", {
    "node_id": "node_abc123",
    "metadata": {
        "status": "archived",
        "archived_at": "2026-03-15T14:30:00Z"
    }
})

# Exclude archived from queries
client.call_tool("memory_recall", {
    "query": "project status",
    "metadata_filter": {
        "status": {"$ne": "archived"}
    }
})

This preserves history and relationships while hiding the node from normal queries.