▦
Centralized compute
Hyperscale data centers increasingly control access to the GPUs that power modern intelligence.
Local-first decentralized AI
The People's AI Network.
Not Big AI.
Koinos AI gives people and developers a path out of centralized AI dependency: run AI on hardware you own, keep sensitive work private, and scale into community compute when you need more.
Built around open AI infrastructure
OpenAI-compatibleLocal-firstSelf-hostedKoinos-powered
K
Your hardware first
Your data stays private
Community compute when needed
Break the dependency on Big AI.Take back control of your data, your compute, your models, and your future.
No permanent vendor lock-inCommunity-owned capacityLocal privacy first
The problem
AI is being concentrated in a handful of data centers.
Today's dominant AI model puts the APIs, hardware, pricing power, and enormous amounts of user data under the control of a few corporations.
⌾
Your data, their infrastructure
Private conversations, files, source code, and business context routinely leave hardware you control.
$
Permanent API rent
Developers build core products around external APIs whose prices, limits, and rules can change at any time.
◇
Innovation behind gates
When access to intelligence is permissioned by a few companies, everyone else builds on rented ground.
Big AI wasn't built to maximize user ownership.Koinos AI is building an alternative.
Our solution
A decentralized AI infrastructure layer.
Use your own machine first. Add trusted private hardware when you have it. When you need more capacity, tap into a distributed network of compatible providers rather than defaulting to a hyperscaler.
Run useful AI locally without network fees
Keep sensitive workloads on hardware you control
Scale into decentralized compute when needed
Reward useful compute providers with KAI
Koinos AI Network
You
▣ Home GPU
▧ Workstation
▤ Local Server
K
Community compute
▦ Providers
◈ Private Pools
⌁ Pro Clusters
LocalYour hardware first
0Required KOIN to start earning
YoursData & routing policy
GlobalCommunity compute
KAIEconomic settlement
Developers
Self-host your AI. Scale with the network.
Keep the familiar API surface while moving control of compute, routing, privacy, and economics back into infrastructure you own.
Your endpoint. Your GPU.
Run an OpenAI-compatible Koinos AI gateway on your workstation, server, or private infrastructure.
app.pylocal-first
client = OpenAI( base_url="http://localhost:8181/v1", api_key="kai_local_••••" ) response = client.chat.completions.create( model="koinos-smart", extra_body={"routing": "local-first"} )
Overflow without lock-in.
Your GPUs serve requests first. If your capacity is exhausted, your routing policy can selectively use private pools or the decentralized public network.
Local-only mode
Private-pool fallback
Public-network overflow
Cost & privacy controls
One API across compute sources.
Apps do not need different architectures for local inference and elastic network inference. Koinos AI handles the routing layer underneath.
Streaming
Project API keys
Usage budgets
Exact model pinning
Earn with your computer
Today1,248.75 KAI≈ $249.75 estimated reference value
AI jobs2,418
GPU time18h 27m
Power estimate$18.64
ModeBalanced
Distribute the infrastructure
Put AI capacity in people's hands.
Instead of requiring every GPU to live in a giant data center, Koinos AI lets compatible consumer and professional computers contribute useful AI work. The software handles benchmarking, model setup, recommended pricing, and resource protection automatically.
01
InstallKoinos AI detects and configures compatible hardware.
02
Start earningNo KAI, KOIN, MANA, or manual pricing required.
03
Do useful AI workVerified compute earns KAI—not passive uptime.
Open compute
From homes to clusters. One network.
Schedulers match workloads with eligible hardware using model availability, performance, reputation, privacy requirements, latency, and economics. Koinos handles ownership and settlement beneath the experience.
⌂
Consumer PCs
Gaming PCs and workstations can contribute capacity when their owners are not using it.
▣
Private pools
Organizations combine hardware they trust and selectively decide whether public overflow is permitted.
▦
Professional providers
Server and GPU fleet operators connect through the same protocol with advanced controls.
◎
Future training
Inference comes first, followed by evaluation, fine-tuning, distillation, and increasingly capable distributed training.
One runtime · two networks
Earn KAI. Secure Koinos. Earn KOIN.
Koinos AI Core can do more than provide AI compute. An optional Koinos node module lets the same machine support the blockchain underneath the AI economy—and advanced users can opt into Proof-of-Burn block production as a separate KOIN earning opportunity.
Share AI capacity.
Let Koinos AI automatically benchmark compatible hardware, route useful AI jobs, protect foreground use, and settle verified work through KAI.
Native rewardKAI
Produce blocks.
Install the optional full-node module and, if you choose, configure Koinos Proof-of-Burn block production. Block production remains governed by Koinos and requires KOIN/VHP.
Native rewardKOIN
Full NodeSupport decentralized chain access and reduce reliance on centralized RPC infrastructure.
Block ProducerAdvanced opt-in consensus participation with separate KOIN economics.
Infrastructure Provider · FutureMeasurable RPC or protocol services used by Koinos AI may qualify for KAI—not passive node uptime.
Clean incentive design: KAI rewards useful AI and Koinos-AI infrastructure services. KOIN rewards Koinos consensus. We do not double-pay KAI simply for producing blocks.
KAI
Value flows to useful AI work.
KAI is designed to coordinate the network economy—not meter AI running on your own hardware.
Use network intelligence.
Pay for third-party inference, verification, future training, specialized models, and other network services.
Earn for useful compute.
Providers receive KAI for successfully completed work rather than simply leaving a process online.
Reward intelligence.
Verified model creators may eventually receive bounded usage royalties when their models serve paid network workloads.
Token supply, fee splits, emissions, bootstrap pricing, and liquidity parameters remain provisional pending economic simulation and legal review.
Privacy by architecture
You decide where the work runs.
Decentralized compute is not automatically private. Koinos AI makes execution policy explicit instead of hiding that distinction.
Your machine. Your model. Data stays on the device.
Use compute owned or explicitly trusted by your organization.
Encrypted transport to reputation-bearing public providers.
Future attested confidential-compute hardware for higher-assurance workloads.
Roadmap
From local AI to distributed intelligence.
The first objective is a working local + network compute loop. More ambitious decentralization and training layers come after the core network proves itself.
Network Genesis
- Local AI & desktop chat
- Provider earning
- KAI settlement
- Self-hosted API
- Network overflow
Open Compute
- Public providers
- More models & hardware
- Private compute pools
- Batch / spot compute
- Model marketplace
Autonomous AI
- Persistent agents
- Tools & permissions
- Machine payments
- Confidential compute
- Enterprise infrastructure
Distributed Intelligence
- Synthetic data
- Fine-tuning
- Distillation
- Distributed training
- Koinos-native models
FAQ
Built to be understandable.
Do I need cryptocurrency to use Koinos AI?
No. Local AI is designed to run on your own hardware without requiring KAI. Network services can be abstracted behind familiar payment and credit experiences.
Do I need KAI or KOIN to start earning?
No. Entry-level providers should be able to begin with compatible hardware and sponsored Koinos resource usage. Higher-value provider classes may later use KAI collateral.
Is decentralized network inference private?
Not automatically. Standard remote workers process the workload. Local, private-pool, and future confidential-compute modes provide stronger privacy choices.
Why is the self-hosted API important?
It lets developers keep their own GPUs as the primary inference layer while retaining an elastic path to decentralized capacity instead of building the product permanently around a hyperscale AI API.
Take back the infrastructure
Own your AI.
Share the scale.
Koinos AI is being built around a simple idea: intelligence should not have to live inside a handful of corporate data centers.