Skip to main content
Agent Circle’s infrastructure is managed hosting all the way down. There’s no Kubernetes cluster to maintain, no bare-metal provisioning, and no dedicated infrastructure team. The goal is to keep operational overhead as close to zero as possible — so that engineering time and budget flow toward product and contributor payouts rather than keeping the lights on.

​
Hosting: Railway and Fly.io

The backend and database run on Railway or Fly.io — both are mainstream managed platforms that handle provisioning, scaling, and deployment without requiring DevOps specialization. Both platforms offer:
  • Managed PostgreSQL. No database administration overhead. Backups, failover, and connection pooling are handled by the platform.
  • Simple deployment pipelines. Push-to-deploy from a Git branch. No custom CI/CD configuration required to get a service running.
  • Pay-as-you-scale pricing. Costs grow proportionally with usage — there’s no large fixed infrastructure commitment in the early phases.
Railway is the current default; Fly.io is a straightforward alternative with comparable managed service coverage and better low-latency edge compute if that becomes relevant in later phases.

​
Cloudflare

Cloudflare sits in front of both the frontend and backend:
  • CDN. Static assets and cacheable API responses are served from Cloudflare’s global edge network, reducing latency for users across regions.
  • DDoS protection. Volumetric and application-layer attacks are absorbed at the edge before reaching the origin.
  • Edge caching for leaderboard data. High-read endpoints like leaderboard rankings are cached at the edge and invalidated on a short TTL — this dramatically reduces backend load under traffic spikes without adding architectural complexity.

​
Frontend

The frontend is a React + Vite static site. Static sites are deployable on any static host — Railway, Cloudflare Pages, Vercel, or any CDN-backed storage bucket. There’s no server-side rendering requirement, which keeps deployment options flexible and hosting costs near zero.

​
Infrastructure Costs by Phase

Infrastructure is kept deliberately inexpensive. The majority of the platform’s operating cost at every phase is engineering time — not servers. This is what makes the developer revenue-share model structurally sound: there’s no large fixed infrastructure obligation that must be covered before payouts become viable. Audit costs in Phase 2 are one-time per program scope, not recurring. Once a program passes audit and is deployed to mainnet, the cost is settled.
Infrastructure costs are published transparently. Contributors can see exactly what their revenue share is funding — there’s no opaque overhead line. When the platform publishes its financials, infrastructure costs are a line item, not an abstraction.

​
Scaling Philosophy

The infrastructure scales with usage and revenue rather than being provisioned ahead of demand. Phase 0 and Phase 1 are intentionally lean — the platform doesn’t need enterprise-grade infrastructure to serve early users, and over-provisioning would reduce the budget available for development and contributor payouts. Phase 3 infrastructure grows in proportion to SDK adoption and on-chain volume. By that point, revenue from agent deployment fees, revenue share, and token activity provides a predictable funding base for infrastructure scaling decisions.