memory_store

Store a new memory node in the knowledge graph.

Parameters

The memory content to store. Should be a clear, complete statement that can be understood without context.

Category of the memory node. You define your own types — common examples: person, project, decision, fact, preference, event.

Additional key-value pairs for structured data. Keys should be strings; values can be strings, numbers, booleans, or arrays.

Array of existing node IDs to create edges to. Creates undirected edges from the new node to each specified node.

Target graph for the node. Defaults to your account’s default graph.

Example request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "memory_store",
    "arguments": {
      "content": "Sarah Chen is the tech lead for Project Phoenix",
      "node_type": "assignment",
      "metadata": {
        "person": "Sarah Chen",
        "project": "Phoenix",
        "role": "tech_lead",
        "start_date": "2026-01"
      },
      "relationships": ["node_sarah_abc123", "node_phoenix_def456"]
    }
  }
}

Example response

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "id": "node_xyz789",
    "status": "created",
    "created_at": "2026-03-03T14:30:00Z",
    "edges_created": 2
  }
}

Response fields

Field	Type	Description
`id`	string	Unique identifier for the created node
`status`	string	Always `"created"` on success
`created_at`	string	ISO 8601 timestamp
`edges_created`	number	Count of edges created via `relationships`

Errors

Code	Description
`INVALID_PARAMETER`	Missing `content` or invalid parameter type
`NODE_NOT_FOUND`	A node ID in `relationships` doesn’t exist
`NODE_LIMIT_EXCEEDED`	Account has reached node limit
`GRAPH_NOT_FOUND`	Specified `graph_id` doesn’t exist

Notes

Free operation — Creating nodes doesn’t count against your query limit.

Write complete content — Nodes should be understandable without relationships. Instead of “She prefers Slack”, write “Sarah Chen prefers async communication via Slack”.

Avoid duplicates — memory_store always creates a new node. Check for existing memories with memory_recall first if you want to update instead.

Hyperedges

To create a hyperedge (multi-way relationship), include the hyperedge parameter:

{
  "arguments": {
    "content": "Q3 planning decision",
    "node_type": "decision",
    "hyperedge": {
      "nodes": ["node_sarah", "node_marcus", "node_phoenix", "node_q3"],
      "edge_type": "decision_context"
    }
  }
}

This creates a new node AND a hyperedge connecting it with all specified nodes.

Idempotency

memory_store is not idempotent — calling it twice with identical arguments creates two separate nodes. If you need idempotent behavior, implement client-side deduplication using content hashing.