graphsub.com / wiki / agent-reliability-graph
Agent Reliability Graph
An agent reliability graph helps teams understand failures by connecting tasks, tools, memory changes, policy gates, and eval results.
Definition
An agent reliability graph maps every meaningful part of an AI workflow: request, plan, tool calls, memory changes, approvals, errors, evals, and final output.
Workflow
Teams can replay failed tasks, cluster failure modes, and fix repeated errors through memory and approval trace improvements.
GraphSub capsule
The Agent Reliability Graph capsule packages this workflow for AI platform teams moving from demos to production operations.
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
- What problems does it solve?
- It helps teams debug tool use, repeated failures, missing approvals, memory drift, and eval regressions.
- Who uses it?
- AI platform teams, automation teams, enterprise copilots, and agent infrastructure teams.