SatsGrid
Permissionless Open-Model Inference, Paid in Sats
About SatsGrid
SatsGrid is an open, permissionless AI inference network that connects independent compute operators with developers and AI agents seeking access to open-weight AI models.
It enables anyone with supported hardware and an open-weight model to operate an independent inference node and monetize available computing resources.
Developers and autonomous agents can discover available inference nodes through Nostr and access models through familiar OpenAI-compatible APIs, paying per request or per token using Bitcoin Lightning.
With no mandatory centralized accounts, subscriptions or API keys, SatsGrid provides an alternative way to access AI inference through permissionless infrastructure and Bitcoin-native payments.
What inspired us?
Access to powerful AI models is becoming increasingly important for developers, researchers, startups and independent builders.
However, accessing the computational resources needed to run these models can be expensive. Many developers depend on centralized AI providers, costly cloud infrastructure or limited free inference services.
At the same time, capable hardware exists around the world outside major centralized cloud providers.
We asked a simple question:
What if anyone could provide AI inference using their own hardware, and anyone could access those resources and pay only for what they use?
That question inspired SatsGrid.
We believe AI infrastructure should be more open, accessible and permissionless. Independent compute operators should have the opportunity to earn from their hardware, while developers should have more freedom in choosing where and how they access AI services.
How it works
SatsGrid brings together three important components.
1. Independent inference nodes
Anyone with supported hardware can operate a SatsGrid node and serve open-weight AI models using inference engines such as Ollama or other OpenAI-compatible runtimes.
Operators configure their models, pricing and Lightning payment backend, creating independent paid inference endpoints.
2. Permissionless discovery
SatsGrid uses Nostr relays for decentralized node discovery.
Operators can announce available models and node information, while developers and AI agents can discover and connect to available inference services without depending on a single central registry.
Access is available through a command-line interface, Rust SDK and OpenAI-compatible API.
3. Bitcoin Lightning payments
SatsGrid integrates the Lightning Network through L402, enabling payment-gated access to AI inference.
Users can pay per request or per token using satoshis, while session budgets provide spending controls for multi-step agent workflows.
This creates a foundation for machine-to-machine commerce, where AI applications can programmatically access and pay for computational resources.
Introducing PayGuard
AI services need more than payment access. They also need predictable and protected payment workflows.
We implemented PayGuard as an integrated payment-integrity capability within SatsGrid.
PayGuard is designed to connect an approved inference request with its provider, model, agreed price and payment conditions.
Its planned capabilities include:
- Request-level payment policies.
- Spending and usage limits.
- Invoice and payment validation.
- Request and payment-state tracking.
- Duplicate-request and replay protection.
- Better handling of unsuccessful inference requests.
- Payment and inference receipts.
We are also investigating service-completion-aware Lightning settlement, allowing qualifying failed requests to be handled before payment is finalized.
Built for an open ecosystem
SatsGrid is designed around open protocols and familiar developer tools.
Its technology stack includes Rust, Bitcoin Lightning, L402, Nostr and open-weight inference engines.
We aim to maintain compatibility with existing OpenAI-style development workflows so developers can integrate SatsGrid into applications and autonomous agents without adopting an entirely new AI development ecosystem.
Compute operators retain control over their hardware, models and infrastructure.
Our goal is to create an ecosystem that developers and operators can inspect, operate and extend.
What comes next?
Our immediate priorities are to strengthen PayGuard, validate the complete Lightning payment lifecycle, improve node discovery and make independent inference services easier to operate and consume.
We envision a growing network of independent compute providers serving developers, AI applications and autonomous agents around the world.
SatsGrid: Open compute. Permissionless access. Payments powered by Lightning.
