Roadmap · Building in public

From gateway to open inference marketplace.

ForgeRouter is being built step by step: first the OpenAI-compatible gateway and ForgeSoftware-hosted models, then ForgeNode onboarding, founding providers, marketplace routing, provider tiers and advanced policies.

OpenAI-compatible API ForgeNode onboarding Community providers Provider tiers Transparent pricing
In progress
Planned
Later
Research
Not committed
In progress

Phase 1: Gateway foundation

Establishing core infrastructure for secure, low-latency API access.

  • check User accounts & API keys
  • check v1/chat/completions endpoint
  • check Prepaid wallet system
In progress

Phase 2: ForgeSoftware provider

First platform-operated deployment to validate routing metrics.

deployment_id: fs-qwen-fr
Model: Qwen3.6-27B
Hardware: 2x A5000
Region: France (EU)
Planned

Phase 3: ForgeNode alpha

Standardized CLI onboarding for secure node enrollment.

$ pipx install forgenode
$ forgenode enroll
$ forgenode detect
$ forgenode publish
Planned

Phase 4: Founding providers

Inviting trusted community members to host models on the network.

Planned

Phase 5: Marketplace beta

Public deployment cards and advanced routing filters.

{
  "model": "meta-llama/Llama-3-70b-chat-hf",
  "routing_strategy": "latency_optimized",
  "filters": {
    "provider_tier": "founding",
    "max_latency_ms": 150
  }
}
Later

Phase 6: Open access

Any provider can join, subject to automated network verification and SLA monitoring.

Research

Phase 7: Advanced policies

Smart routing based on cost, data privacy regions, or hardware specific requests.

{
  "routing_strategy": "cost_optimized",
  "filters": {
    "region": "EU_ONLY",
    "min_vram_gb": 40
  }
}
Not committed

Phase 8: Decentralized settlement

Exploring trustless payment channels for provider settlements.

What we are not promising yet.

block No fake SLAs
visibility_off No hidden trust downgrades
money_off No guaranteed provider revenue

What we want to validate first.

speed Gateway latency overhead under load
security Secure isolation for ForgeNode hosts
analytics Accuracy of empirical throughput metrics

How to participate

code

For Developers

  • checkAccess multiple providers through a single API.
  • checkControl routing based on empirical metrics.
  • checkUnified billing through a prepaid wallet.
Read API Docs
dns

For Providers

  • checkMonetize idle GPU capacity easily.
  • checkStandardized onboarding via CLI.
  • checkTransparent performance metrics.
Apply as Founder

Build log

Oct 2023

Gateway Alpha launched

Internal testing of core routing layer.

Sep 2023

Initial concept

Architecture design and technical validation.

FAQ

How is routing latency measured? expand_more

We maintain active probing connections from our gateway to provider nodes, measuring Time To First Token (TTFT) continuously to inform our routing tables.

Can I bring my own models? expand_more

During early phases, we restrict the catalog to verified popular open weights. Custom model hosting is planned for Phase 6.

Build the open inference marketplace with us.

Join the private beta arrow_forward Become a provider