Pilot program for engineering teams

Reduce AI coding spend without changing developer workflow.

FanOn routes requests between local models and cloud models, keeping simple work local and escalating harder work when needed.

Join Pilot Read trust model

OpenAI-compatible Local-first routing Provider fallback No prompt storage by default

local_first policy

IDE / AI client

FanOn router

Local worker

Cloud fallback

68% local route target

32% provider route target

Example dashboard values for design direction. Real metrics depend on pilot traffic.

Problem

AI coding tools are useful, but provider spend is becoming hard to ignore.

Costs scale with every developer

IDE assistants make teams faster, but cloud-model usage can become a growing recurring infrastructure cost.

Workflow changes hurt adoption

Engineers should not need to leave Cursor, Continue, Cline, or other tools just because the organization wants better routing.

Local capacity is underused

Laptops, workstations, and shared machines can handle many routine AI tasks, but teams need a transparent routing layer to use them well.

How FanOn Works

A routing layer between AI clients and execution backends.

FanOn exposes an OpenAI-compatible endpoint. Existing clients point at FanOn, and FanOn routes each request through local workers first, with explicit provider fallback when configured.

Connect existing AI clients

Use OpenAI-compatible settings from IDE tools and local playgrounds.

Route local when possible

Routine, private, or low-cost work can stay on local or team-owned machines.

Escalate when needed

Provider fallback remains available for harder requests or unavailable local targets.

Benefits

Built for pilots that measure cost reduction and developer experience together.

Lower provider usage

Track local versus provider execution and estimated avoided spend.

Keep developer workflow

Start with OpenAI-compatible IDE workflows instead of a new app mandate.

Visible routing decisions

Inspect route history, fallback reasons, latency, and target selection.

Privacy-aware by default

FanOn is designed around no prompt storage by default and local-first execution.

Privacy & Trust

Optimization infrastructure, not employee monitoring.

FanOn focuses on aggregate routing, cost, latency, and reliability signals. The product direction explicitly avoids prompt inspection, productivity scoring, and manager visibility into individual conversations.

No prompt storage by default
No employee productivity scoring
Transparent route decisions
Aggregate team metrics before individual metrics

Pilot Dashboard

See what value FanOn would surface after installation.

Example pilot dashboard Illustrative metrics from a 7-day local/dev pilot shape. Real numbers depend on team traffic, configured workers, models, and provider fallback policy.

Value Overview 7 days

68% worker/local

32% provider

Estimated spend avoided $418

Estimated savings 41%

Value overview

Shows the local/provider split and the estimated provider spend avoided.

Routing Activity latest

local mock-code worker-a

local qwen2.5-coder worker-b

fallback gpt-4o-mini local unavailable

local mock-code worker-a

Routing activity

Shows selected targets, fallback reasons, and recent routing decisions.

Topology local/dev

FanOn API

worker-a
healthy

worker-b
healthy

provider
explicit fallback

Topology

Shows live workers, provider fallback, and the local/dev control plane.

Pilot Program

Designed for 5-10 engineer teams evaluating AI coding cost optimization.

A FanOn pilot tests whether local-first routing can reduce provider usage while keeping latency, reliability, and developer satisfaction within acceptable bounds.

2-4 week pilot window Existing IDE workflow first Local workers plus optional provider fallback Aggregate metrics and trust boundaries

FAQ

Common pilot questions.

Is FanOn production ready?

No. FanOn is currently a local/dev MVP for pilots and design partner discovery.

Does FanOn store prompts?

No prompt storage by default is a core trust principle. Current metrics avoid prompt and message content.

Can engineers keep using their IDE assistants?

That is the intended near-term wedge: OpenAI-compatible tools should point to FanOn with minimal workflow changes.

Does FanOn replace cloud models?

No. FanOn is a routing and optimization layer. Premium provider models remain useful for harder requests and fallback.

Join Pilot

Help shape FanOn with a real engineering workflow.

Tell us about your team, AI tooling, and cost pressure. The current pilot focus is engineering organizations with active AI coding assistant usage.