AI agents,
powered by sunlight.
OpenAI-compatible API across 50+ models. Persistent agents that remember across sessions. GPU compute that follows the sun — running on 100GW of planned solar capacity in Argentina.
Three layers.
One platform.
Each layer is independently powerful. Together they form a complete stack — from multi-engine orchestration down to persistent agent cognition.
XCT-Builder
Multi-engine routing across LangGraph, CrewAI, n8n, Dify — with MCP-standardized context.
XCT-Exchange
Four-factor bus connecting Model, Data, Compute and Energy hubs through a single rail.
XCT-Agent
Persistent memory and cross-model sessions — agents that remember across time and models.
Four factors.
One bus.
Model · Data · Compute · Energy — connected through a unified rail. Workloads are routed to the right resource at the right price, settled on solar.
- Model Hub 50+ models
- Data Hub Ocean C2D
- Compute Hub GPU clusters
- Energy Hub 100GW solar
Routed by code.
Settled on sun.
Every task on Xcity is routed by an engine, executed on GPU clusters and metered against real solar generation — visible end-to-end through a single CLI and API.
Programmable
HTTP, WebSocket and CLI surface — built for agents and humans alike.
Deterministic Memory
Vector + structured recall, replayable across sessions and models.
Solar Settled
Compute clears against on-chain solar tokens — not abstract credits.
Human in Loop
Reviewable checkpoints for high-stakes decisions, on by default.
Solar capacity, programmable as compute.
Xcity's energy backbone is anchored in Argentina — 2GW operational, a 100GW pipeline. Every kilowatt becomes addressable infrastructure.
Solar ComputeWhat teams are saying
"We migrated our entire inference stack to Xcity's TokenHub in a weekend. 50+ models behind one API key, with per-team budget controls. Our compute cost dropped 30% on day one."
"Hermes Agent is the first AI that actually remembers my business context across sessions — pricing strategy, codebase quirks, team conventions. It's like a brilliant collaborator who never needs onboarding."
"The Argentina solar compute infrastructure is a genuine moat. We locked in reserved GPU capacity at a fraction of hyperscaler pricing. Latency is better than we expected and the uptime has been flawless."
Get early access.
Before the crowd.
Drop your email — we'll reach out when API access opens.