AI Optimizer ← Back to home
Install guide

Download it, start it, and test it in a few minutes.

If the release email brought you straight to GitHub, this is the cleaner setup path. Download the latest build, activate your license, start the local proxy, and run one quick check against localhost:3000.

Quick answer: AI Optimizer is a local desktop app that runs a proxy on http://localhost:3000. Start the free trial first, then download the app, activate your license, save your provider API key, and point your tools at the local endpoint.
1

Download the current release

Get the latest macOS, Windows, or Linux build from the GitHub Releases page.

2

Install the app for your platform

Use the macOS zip, Windows installer or packaged executable, or the Linux AppImage / deb package depending on your machine.

3

Activate your license

Open AI Optimizer, paste your license key, click Activate, and confirm the license is active before starting the proxy.

4

Choose a provider and save the key

Select OpenAI, Anthropic, or Google Gemini in the app, then save the API key for the provider you want to run.

5

Start the local proxy

In the Proxy Server section, click Start. The app will run locally on http://localhost:3000.

6

Run one quick test

Use the health and chat checks below to confirm the proxy is live and a real request succeeds.

AI Optimizer desktop app showing provider settings, proxy controls, and cache stats
A good first-run state: provider selected, key saved, proxy running, and stats visible in one place.
macOS

Install and open

Download the macOS zip from Releases, unzip it, and move AI Optimizer.app into Applications.

xattr -r -d com.apple.quarantine /Applications/AI\ Optimizer.app

If macOS blocks launch on first open, removing quarantine is usually enough.

Windows

Install and allow the local proxy

Run the Windows installer or packaged executable from Releases, then launch AI Optimizer.

Windows Defender Firewall may ask for permission the first time you launch it. In most cases, allowing Private networks is enough because the app runs a local proxy on localhost:3000.

Linux AppImage

Make it executable and run it

chmod +x AI\ Optimizer-2.4.0.AppImage
./AI\ Optimizer-2.4.0.AppImage
Linux deb

Install with apt

sudo apt install ./ai-optimizer_2.4.0_amd64.deb
Point your tools at the proxy

Use the local endpoint instead of the upstream base URL.

For many scripts, tools, and local agent workflows, the practical setup change is just routing requests through AI Optimizer first.

OPENAI_BASE_URL=http://localhost:3000/v1

The same local-first pattern applies when you are using AI Optimizer with Anthropic or Google Gemini through the supported app lanes.

Test 1

Health check

curl -sS http://localhost:3000/health

Expected response: {"status":"ok","running":true}

Test 2

Stats check

curl -sS http://localhost:3000/stats

Use this to confirm requests, exact cache hits, and OpenAI partial prompt reuse stats.

macOS / Linux

Quick OpenAI chat test

curl -sS http://localhost:3000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"Reply with exactly: INSTALL_OK"}]}'
Windows CMD

Quick OpenAI chat test

curl -X POST http://localhost:3000/v1/chat/completions -H "Content-Type: application/json" -d "{\"model\":\"gpt-4o-mini\",\"messages\":[{\"role\":\"user\",\"content\":\"Reply with exactly: INSTALL_OK\"}]}"
What to expect

Exact hits vs partial hits

Exact Cache Hits mean the repeated request was fully served from local cache. Partial Hits (OpenAI) only move when OpenAI reports real provider-side reused prompt tokens.

Current scope

Provider support stays honest

OpenAI remains the broadest lane. Anthropic support is focused on chat completions. Google Gemini support is narrower in this app lane. One active provider at a time keeps the workflow simple.

Keep the workflow. Cut the repeated waste.

Start the free trial to get your license, then download AI Optimizer and point your tools at the local proxy.

Start free trial Download latest release