swarmagents.codes

Swarm Topology & Task DAG Graphquantum_router.py

Active Nodes: 10 · DAG Depth: 3 · Polling /api/v1/state

Mini Map
System TPS1840
Avg Latency24 ms
GPU Load42%
Quantum Router Pulses14290

Quantum Router

quantum_router
Status
firing
Active Tasks
2
Response Latency
8 ms
Compute Load
88%
Task DAG Node MetadataDepth L0
Action:decompose_task
Quantum Confidence:94%
Quantum Router EngineQRCBandit 27D
Epsilon
0.05
Success Rate
96%
Active Action
decompose_task
Top Routing ProbabilitiesDim: 27→30
decompose_task
42%
execute_task
28%
validate_output
15%
Message Pings Stream/api/v1/chat
No message pings captured yet. Dispatch a task below.
Command center online

A company of agents
on your models.

Decompose thousand-step goals into parallel execution chains, spawn seats that invent their own roles, and watch the swarm ship — from one glass cockpit.

Watch a live swarm
Runs on your providers 128 seats warm Pure seat spawner

Token burn rate

$142.37/hr

Tokens streamed

4,812,940

since session start

Global latency

212 ms

· p50 within SLO

Fleet

seats128/160
blackboard sync99.98%
Telemetry

Swarm CLI

Drive the whole org
from one prompt line.

Autocomplete across every swarm primitive. Decompose, spawn, cap burn, deploy — the CLI compiles intent into seat schedules over live SSE streams.

CLI reference
swarm-cli — liveSSE

swarm v2.4.1 — 128 seats online · shared blackboard synced

decomp engine

Decompose anything

Time-scaled trees deepen until work is leaf-sized — then seats fan out and execute in parallel.

seat spawner

Spawn a real team

Dynamic seats invent their own roles, own paths through the repo, and coordinate like a company.

shared memory

Share one workspace

Blackboard, heatmap, and live SSE context keep every agent oriented without you babysitting.

Zero-config · bring your own keys

Point the swarm at a repo.
Watch it work.

One OAuth click. Your code stays in your org; the seats do the typing.