Execution-first
Every feature — spatial routing, circuit breakers, dead-letter queues — is designed around getting agent work done under real latency and quota constraints.
Scale Systems Docs
Documentation
Platform Overview
Why Scale Systems AI
Scale Systems AI is the distributed execution and telemetry engine for multi-agent swarms. We give teams a single control plane to provision agents, route workloads across spatial nodes, meter compute, and recover from failure — without rebuilding orchestration from scratch.
Distributed swarm runtime · Production ready
Our mission
Autonomous agents are no longer experiments — they are production workloads. Our mission is to make multi-agent systems as reliable, observable, and billable as any other cloud service. Scale Systems turns swarms of specialized agents into a managed execution fabric: you define objectives, we handle dispatch, telemetry, retries, and settlement across every node in the fleet.
Telemetry-native
Live SSE feeds, webhook delivery, and SRE dashboards are built into the platform core, not bolted on after launch.
What makes Scale Systems different
Most agent frameworks stop at the prompt loop. Scale Systems extends through the full operational stack: provisioning, spatial persistence, provider health tracking, model fallback routing, and dual-rail billing on a unified entitlement model.
Spatial swarm orchestration
Agents are placed in a 3D spatial model where proximity, affinity, and load shape dispatch — not just FIFO queues.
Self-healing execution
Circuit breakers, provider health scores, and automatic model fallback keep swarms running when upstream APIs degrade.
Dual-rail billing
Stripe and PayFast rails share one quota model — tokens, agents, and tier gates apply identically across fiat settlement paths.
Workspace RBAC
OWNER, ADMIN, OPERATOR, and VIEWER roles gate every control-plane API and dashboard surface for enterprise tenancy.
Who uses Scale Systems
Teams that run agent fleets in production — not one-off chatbots. Typical adopters include platform engineering groups building internal automation, AI product teams shipping swarm templates to customers, and SRE organizations that need measurable reliability around autonomous workloads.
Platform teams
Internal developer platforms that need a governed agent runtime with API keys, quotas, and audit trails.
AI product builders
Teams shipping lead-sentinel, content, and ops swarms from templates with metered usage and customer-facing telemetry.
SRE & ops
Operators who monitor provider health, chaos scenarios, and dead-letter recovery without SSHing into worker nodes.
Built to scale
From a single sandbox agent to unlimited Premium-tier throughput, the platform scales horizontally. BullMQ workers, Redis-backed queues, and Prisma-backed persistence let you add execution nodes without rewriting orchestration logic.
FREE
1 agent
50K tokens / month
STARTER
5 agents
500K tokens / month
PREMIUM
Unlimited
No throughput ceiling
Architecture
A control plane (Next.js API routes + workspace RBAC) sits above distributed workers that execute swarm graphs. Spatial persistence tracks node placement; provider health and circuit breakers gate model calls; telemetry streams outward via SSE and signed webhooks.
Control plane
- REST API v1 with tier rate limits
- Workspace-scoped RBAC on every route
- Agent provisioning and API key vault (SHA-256 hashed)
- Dual-rail billing and entitlement sync
Execution plane
- BullMQ workers for swarm node execution
- Self-healing executor with dead-letter handling
- Model fallback router and provider health tracker
- Spatial swarm persistence and 3D telemetry UI
Deep integration reference: Developer Guide · API Reference