Temporal Memory

Every node in HyperMemory knows when it was created and when it was last modified. This temporal awareness lets your agent reason about knowledge across time.

Every node knows when

All nodes automatically track:

Field	Description
created_at	When the memory was first stored
updated_at	When it was last modified

{
  "id": "node_abc123",
  "content": "Our main competitor is Acme Corp",
  "node_type": "fact",
  "created_at": "2026-01-15T10:30:00Z",
  "updated_at": "2026-02-28T14:22:00Z"
}

Why temporal memory matters

Avoiding stale knowledge

Without temporal context, your agent might act on outdated information:

Agent: "Our main competitor is Acme Corp"
Reality: Acme Corp was acquired 3 months ago

With temporal queries, your agent can ask: “What do we know about competitors that’s been updated in the last month?”

Understanding evolution

Knowledge changes over time:

January:  "API response time is 200ms"
February: "API response time is 150ms"  (after optimization)
March:    "API response time is 100ms"  (after caching)

Temporal queries let you see this progression.

Point-in-time questions

Sometimes you need historical context:

• “What did we know about the competitor in January?”
• “What was our understanding before the Q3 decision?”
• “What changed after the migration?”

Querying temporal data

Filter by time range

Use the time_range parameter to filter results:

memory_recall(
  query="What do we know about API performance?",
  time_range={
    "start": "2026-02-01T00:00:00Z",
    "end": "2026-02-28T23:59:59Z"
  }
)

This returns only memories created or updated in February 2026.

Get recent memories

memory_recall(
  query="What happened this week?",
  time_range={
    "start": "2026-03-01T00:00:00Z"
  }
)

Omitting end means “up to now.”

Get historical snapshot

memory_get_history(
  node_id="node_abc123",
  as_of="2026-01-31T23:59:59Z"
)

Returns what the node looked like at the end of January.

Version history

Nodes maintain version history. When you update a node, the previous state is preserved:

memory_update(
  node_id="node_abc123",
  content="API response time is 100ms after caching improvements"
)

You can retrieve the history:

memory_get_versions(
  node_id="node_abc123"
)

Response:

{
  "versions": [
    {
      "version": 3,
      "content": "API response time is 100ms after caching improvements",
      "updated_at": "2026-03-15T10:00:00Z"
    },
    {
      "version": 2,
      "content": "API response time is 150ms after optimization",
      "updated_at": "2026-02-15T10:00:00Z"
    },
    {
      "version": 1,
      "content": "API response time is 200ms",
      "created_at": "2026-01-15T10:00:00Z"
    }
  ]
}

Common temporal patterns

Freshness filtering

Prioritize recent information:

memory_recall(
  query="team preferences",
  time_range={"start": "2026-01-01T00:00:00Z"},
  sort_by="updated_at",
  sort_order="desc"
)

Change detection

Find what changed since last check:

memory_recall(
  query="project status",
  time_range={
    "start": last_check_timestamp
  }
)

Historical comparison

Compare knowledge at two points:

january_state = memory_recall(
  query="Q1 priorities",
  time_range={"end": "2026-01-31T23:59:59Z"}
)

march_state = memory_recall(
  query="Q1 priorities",
  time_range={"end": "2026-03-31T23:59:59Z"}
)

Best practices

Next steps

MCP Server Overview

Learn how HyperMemory exposes these features

memory_recall API

Full reference for temporal queries