graphsub.com / wiki / agent-memory-graph
Agent Memory Graph
An agent memory graph organizes working, episodic, and semantic memory so autonomous systems can recall, update, and audit context over time.
Definition
An agent memory graph connects short-lived working memory, episodic task history, semantic facts, tool outputs, and user corrections into one queryable structure.
Core model
GraphSub models memory as typed nodes and edges with importance, source, timestamp, scope, and replay metadata. That makes memory inspectable rather than hidden inside a prompt window.
Use cases
Agent memory graphs are useful for customer support agents, research assistants, compliance workflows, code agents, operations copilots, and any system that must remember why it acted.
Code example
Agent memory and replaypython
from graphsub import Client
client = Client(base_url="https://graphsub.com")
memory = client.memory.remember(
tier="working",
content="limit breach on EUR rates book",
source="risk-run-2026-07-03",
importance=0.92,
)
trace = client.capsules.execute(
capsule="trading-control-graph",
workload="risk-recompute",
evidence="source-linked",
)
print(trace.replay_id, trace.affected_outputs)Write agent memory, launch a capsule, and inspect affected outputs.
FAQ
- Why use a graph for agent memory?
- Graphs preserve relationships between tasks, tools, people, documents, decisions, and outputs.
- Can memory be audited?
- Yes. GraphSub memory can be traced through sources, tool calls, approvals, and replay events.