⚡

AI Agent Network Simulator

Topological routing, compute caps & bottleneck optimization

Interactive DAG Infrastructure Map Ready
Total Throughput 0 TPS Target: 250 TPS
Avg Settlement Latency 0 ms Optimal route
Rate Limit Stalls 0 Dropped transactions
Escrow Pool Balance $2,500 API Credits: 5,000
🔍 Graphlib Topological Bottleneck Analysis

Analyzing critical DAG path and rate limit boundaries...

AI Agent Infrastructure Optimization Result Verified State
Simulated Peak TPS:
170 TPS
Critical Bottleneck Node:
agent-beta (DATA_FETCHER)
Settlement Path Status:
Stable (17ms total delay)
Node Inspector agent-alpha
Connect New Channel
Schema Export View
{ "loading": true }

A local topology heuristic finds the smallest assigned node capacity

Read the explanation

Four initial nodes assign capacities fifty, one hundred twenty, two hundred and three hundred transactions per second. Analysis reports the smallest capacity, fifty, by scanning every node rather than calculating path throughput. At one pixel per capacity unit the smallest bar fifty contrasts with the sum six hundred seventy drawn at half scale as three hundred thirty five. Adding a parallel edge does not change that minimum. These figures are local fixture capacity labels, not observed traffic or an executed network benchmark. Four links assign twelve, five, eight and six milliseconds latency. Their sum is thirty one and arithmetic mean seven point seven five, rounded to eight. A path through beta is seventeen milliseconds, while a path through gateway is fourteen, but the display reports average link delay, not either path latency. At ten pixels per millisecond the bars are eighty, one hundred seventy and one hundred forty. Bandwidth and compute labels do not enter this latency or throughput calculation. The stall label is a fixed fourteen per second when minimum capacity is below one hundred, not a queueing estimate. The bottleneck preset assigns minimum thirty and mean link delay twenty two point five, rounded twenty three. The scaled preset assigns minimum three hundred fifty and mean delay three, making a fixture ratio of eleven point six seven for minimum capacity. At one pixel per transaction per second the two capacity bars are thirty and three hundred fifty; their difference three hundred twenty. The embedded graph library fallback works offline, but its DFS ignores cycles rather than rejecting them. A rendered active DAG label therefore does not prove acyclicity. JSON exports the assigned topology and analysis; animated credit jitter is local random decoration, never provider credit consumption.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.