memory_load_link

Import an exported subgraph into your knowledge graph. Use to restore backups, share knowledge between agents, or migrate data.

Parameters

The exported subgraph JSON (from memory_export_subgraph).

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

How to handle imported data. Options:

"append" (default): Create new nodes with new IDs
"replace": Replace existing nodes if IDs match (use with caution)

Prefix to add to all imported node IDs. Useful for namespacing imported data.

Example request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "memory_load_link",
    "arguments": {
      "data": {
        "nodes": [
          {
            "id": "node_abc123",
            "content": "Sarah Chen, Backend Tech Lead",
            "node_type": "person",
            "metadata": {"team": "backend"}
          }
        ],
        "edges": [
          {
            "edge_type": "manages",
            "source": "node_abc123",
            "target": "node_def456"
          }
        ],
        "hyperedges": []
      },
      "graph_id": "production",
      "merge_strategy": "append",
      "id_prefix": "imported_"
    }
  }
}

Example response

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "import_id": "imp_xyz789",
    "imported_at": "2026-03-15T14:30:00Z",
    "target_graph": "production",
    "id_mapping": {
      "node_abc123": "imported_node_abc123",
      "node_def456": "imported_node_def456"
    },
    "stats": {
      "nodes_created": 2,
      "nodes_updated": 0,
      "nodes_skipped": 0,
      "edges_created": 1,
      "hyperedges_created": 0
    },
    "warnings": []
  }
}

Response fields

Field	Type	Description
`import_id`	string	Unique identifier for this import
`imported_at`	string	Import timestamp
`target_graph`	string	Graph data was imported to
`id_mapping`	object	Maps original IDs to new IDs
`stats.nodes_created`	number	New nodes created
`stats.nodes_updated`	number	Existing nodes updated (replace mode)
`stats.nodes_skipped`	number	Nodes skipped due to errors
`stats.edges_created`	number	Edges created
`stats.hyperedges_created`	number	Hyperedges created
`warnings`	array	Non-fatal issues encountered

Errors

Code	Description
`INVALID_PARAMETER`	Missing `data` or invalid structure
`INVALID_EXPORT_FORMAT`	Data doesn’t match export schema
`NODE_LIMIT_EXCEEDED`	Import would exceed node limit
`GRAPH_NOT_FOUND`	Specified `graph_id` doesn’t exist

Notes

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

Requires admin scope — This operation requires memory:admin scope.

Merge strategies

append (default)

All nodes get new IDs
Safe — never overwrites existing data
Use for: Sharing knowledge, adding new data

replace

Nodes with matching IDs are updated
Nodes without matches are created
Use for: Restoring backups, syncing data

Replace mode warning — replace can overwrite existing nodes. Only use when you’re sure you want to overwrite.

ID mapping

Imported nodes receive new IDs. Use the id_mapping response to track:

response = client.call_tool("memory_load_link", {"data": export_data})

# Find the new ID for an imported node
old_id = "node_abc123"
new_id = response["id_mapping"][old_id]

Edge handling

Edges reference node IDs. During import:

Source and target IDs are mapped to new IDs
If both nodes exist (imported or pre-existing), the edge is created
If either node is missing, the edge is skipped (warning added)

Import validation

Before importing, the data is validated:

JSON structure matches export schema
All edge endpoints reference valid node IDs
Hyperedges have at least 3 nodes

Invalid items are skipped with warnings.

Use cases

Restore from backup:

{
  "data": backup_data,
  "merge_strategy": "replace"
}

Share knowledge between agents:

{
  "data": agent_a_export,
  "graph_id": "agent_b_graph",
  "id_prefix": "from_agent_a_"
}

Migrate to new graph:

{
  "data": old_graph_export,
  "graph_id": "new_graph",
  "merge_strategy": "append"
}

Large imports

For large imports (>1000 nodes):

Consider batching into multiple smaller imports
Monitor node limits before importing
Use id_prefix to namespace batches

for batch in chunks(nodes, size=500):
    client.call_tool("memory_load_link", {
        "data": {"nodes": batch, "edges": [], "hyperedges": []},
        "id_prefix": f"batch_{batch_num}_"
    })