Private inference, GPUs and always-on agents on one network. Pay per use in USDC, or bring a GPU and earn.
Inference, machines, agents, processors and data on one account and one USDC balance. Grid inference runs on TEE-attested nodes, sealed end to end.
Decentralized confidential inference. Submit a prompt, the network routes it to an attested node, and the sealed response comes back end-to-end encrypted. Pay per token in USDC, settle on Solana.
On-demand cloud & GPU instances. Deploy a VPS or GPU box in minutes, pay daily in USDC.
Deploy one function, get a paid API and a live MCP server. Buyers pay you in USDC over x402.
One-click, always-on AI agents on dedicated CPU. Your agent gets its own wallet, memory and skills — talk to it on Telegram, Discord or Slack. Bring your own AI key or use ours.
A GPU that boots with an LLM ready to go. Get your own OpenAI-compatible endpoint in minutes — keep it private, or join the grid and earn.
Zero-knowledge encrypted backups of your agents, on pods or your own machines. Restore anywhere, even onto a fresh pod.
Describe the data you need and get a ready-to-use training dataset. Generate privately on the grid, pay per row.
Every node runs an allow-listed binary inside a TEE and proves it before any job routes to it. Tampering is slashable.
Stake $SGL to run nodes or validate. Earn a share of every settled job in USDC + SGL. Slashable only for tampering, never for downtime.
Client-encrypted chat over the grid. Your prompts and history never touch our servers.
OpenAI-compatible endpoints. Agent-purchasable via x402. Bill per request, no account needed.
Pay from Solana or Base in USDC. Credits or x402. No wrapping, no bridges.
No proprietary cloud, no trust in operators. Hardware roots of trust, verified by the network before any job runs.
Your prompt is encrypted client-side and posted to the grid. The orchestrator matches it to a candidate node by model, price and availability.
Each node runs an allow-listed binary inside a TEE and proves it. Only attested nodes ever receive a job.
Inputs are decrypted only inside the enclave. The response is sealed straight back to you. No plaintext touches our servers.
Payment splits on Solana across node, stakers, team and burn. Provably wrong work is slashable.
A pod is a persistent AI agent on its own dedicated CPU machine — with its own crypto wallet, memory and skills. It sits in your Telegram, Discord or Slack. Bring your own AI key or use ours.
The raw agent, built on OpenClaw. Name it, write its personality, install skills, connect channels — it does whatever you build it to do.
The same raw agent on Nous Research's Hermes — pick the brain, keep the body: wallet, memory, channels and skills.
Runs your Telegram community. Answers members, removes scams and impersonators, and sends you a private briefing about what actually happened.
Watches pages, APIs and on-chain addresses you would otherwise keep checking — and messages you only when something actually changes.
Every pod ships its own OpenAI-compatible endpoint. Mint a key in the app and your agent answers inside the tools you already use.
https://compute.x402layer.cc/pods/<your-pod>/v1The fastest path from “I want an LLM” to a running one. Pick a GPU, pick a model that actually fits it, and get your own private endpoint — or serve the grid and earn.
From a small A16 to serious cards. We only show models the card can actually run — real context-window fit, no 14B-on-16GB traps.
Llama, Qwen, Gemma, Mistral, DeepSeek R1 — general, coding or reasoning. Preinstalled and serving the moment the machine boots.
Point any OpenAI client at your private endpoint — even give it a memorable name at ai.x402compute.cc/<your-slug>. Full /v1 surface, /models included, so you can even apply as an OpenRouter provider.
Deploy the same GPU as a grid node: serve the network’s inference and earn USDC + SGL on every job. Needs 50k SGL staked.
General-purpose shared vCPU for services, agents & APIs.
Faster high-frequency CPU for latency-sensitive workloads.
Fully dedicated CPU & RAM for heavy, steady workloads.
NVIDIA GPU instances for inference, training & rendering.
The core is live. These are next.
Confidential, decentralized training & fine-tuning. Run jobs across attested GPU nodes with your data sealed end-to-end.
Co-own a real AI machine with others. Each owner gets exclusive time on it, and a share of the payout when it is resold.
Spin up your own branded slice of the grid. Recruit nodes, route inference, and earn on every request.
Shard one giant model across many grid nodes. Pooled VRAM runs models too large for any single machine to hold.
Own and trade slices of real compute. Clusters launch their own coins, and holders earn from every job their hardware serves.
SGL Cloud is the control room for everything on this page. Chat on the grid, deploy Machines, AI Machines and Agent Pods, run your nodes, and manage credits & API keys — one account, one balance, paid in USDC.
$SGL is the collateral that secures the network and the asset stakers earn from. It is live on Solana.
Turn a Mac or server into a confidential compute node. Stake $SGL, serve models, and earn in USDC + SGL. Single-command install, hardware attestation native.
No terminal required. Pick a model, connect your wallet, press Connect — and start earning.