Become a ForgeRouter provider.

Connect your vLLM, llama.cpp, SGLang or OpenAI-compatible endpoint, publish your models, set your own pricing, and let ForgeRouter handle API access, usage tracking, billing and provider visibility.

Apply as a founding provider See how ForgeNode works
Provider Infra vLLM / llama.cpp ForgeNode Local Gateway ForgeRouter API Gateway, Billing, Usage, Listing Developers API Consumers

Who can become a provider?

Community GPU host

You have a rig with a few GPUs (e.g. RTX 3090/4090s) sitting idle. You want to monetize that compute with minimal effort.

Fine-tune creator

You created a great fine-tune of Llama 3 or Mistral. You want to offer it as an API without building a billing platform from scratch.

AI agency

You host internal models for clients. You have excess capacity during off-peak hours that could generate revenue.

GPU hosting provider

You rent bare-metal GPUs. Offering inference APIs alongside your bare-metal rentals increases your margin and utilization.

Datacenter / Professional

You operate a verified tier datacenter in Europe and want to receive enterprise workloads specifically routed to your region.

You run the model. ForgeRouter handles the marketplace layer.

You provide

  • check Hardware (GPUs, networking, power)
  • check Inference runtime (vLLM, llama.cpp, etc.)
  • check Model weights and quantization
  • check Uptime and availability
Our Job

ForgeRouter handles

  • check Global API Gateway & Request routing
  • check Token accounting and usage tracking
  • check Developer billing and invoicing
  • check Model catalog listing and discoverability
  • check Provider payouts

ForgeNode: from local endpoint to marketplace deployment.

ForgeNode is the lightweight Python agent providers install next to their inference runtime. It securely bridges your local API to the global mesh without complex firewall rules.

memory Detects GPU & VRAM
settings_applications Detects Runtime
psychology Detects Model
monitor_heart Reports Health
terminal 
$ pipx install forgenode
Installed forgenode-0.2.4

$ forgenode enroll --token <YOUR_TOKEN>
Authenticated as org_forge_devs

$ forgenode detect
Found vLLM running on http://localhost:8000
Detected Model: Qwen/Qwen2.5-32B-Instruct
Detected Hardware: 2x NVIDIA RTX A5000 (24GB)

$ forgenode publish
Deployment active. Routing traffic to gateway...

Start open. Build trust over time.

All providers start in the sandbox. Consistently high uptime and fast latency unlock higher tiers and better visibility.

science

Sandbox

Testing integration. No public traffic.

group

Community

Available via economy routing. Best effort.

visibility

Observed

Metrics tracked. Included in balanced routing.

verified

Verified

Identity checked. High trust signals.

domain

Professional

SLAs provided. Enterprise routing ready.

You set your inference price.

Provider inference price + ForgeRouter platform fee = Effective customer cost
  • check You define the cost per 1M input/output tokens.
  • check We take a small platform fee to cover gateway hosting, billing, and bandwidth.
  • check Payouts are processed monthly via Stripe Connect or wire transfer.

Honesty note: During the beta, provider payouts are processed manually at the end of the month. Automated Stripe Connect payouts will be enabled before general availability.

Example Provider Listing
Model: forge-1/Qwen3.6-27B
Provider: Acme Infra
Tier: Observed
Location: Germany
Hardware: 4x RTX 4090
Provider Price: €0.15 / 1M tokens
Platform Fee: €0.05 / 1M tokens
Customer Pays: €0.20 / 1M tokens

We are looking for 5 founding providers.

Join the initial cohort. We are actively onboarding reliable hosts with unique models or competitive pricing to seed the marketplace.

star Benefits

  • • Priority listing in the model catalog
  • • Direct support channel with the founding team
  • • Influence over the ForgeNode roadmap
  • • Zero platform fees for the first 3 months

rule What we ask

  • • Willingness to tolerate early beta bugs
  • • Commitment to running the node for at least 30 days
  • • Clear communication regarding planned downtime

Application Form

We review applications manually and aim to respond within 48 hours.

Frequently Asked Questions

Do I need a static IP?

No. ForgeNode opens a secure, outbound connection to the ForgeRouter gateway. You do not need to open inbound ports or have a static IP, though a stable connection is required.

What happens if my node goes down?

The gateway will detect the failure almost immediately and route subsequent requests to other providers hosting the same model. Frequent downtime will affect your reliability score and tier status.

How do I get paid?

Payouts are calculated based on the tokens you processed successfully. During beta, this is handled manually via invoice/wire transfer. We will implement Stripe Connect for automated payouts soon.

Is the code open source?

The ForgeNode agent you install on your machine will be open source so you can verify exactly what it detects and transmits. The central ForgeRouter gateway and billing systems are proprietary.