linshenkx/prompt-optimizer

An AI prompt optimizer for writing better prompts and getting better AI results.

Prompt Optimizer – What It Is

Prompt Optimizer is a web‑based (and also desktop / Chrome‑extension / Docker) tool that helps you improve the prompts you send to large language models (LLMs) and text‑to‑image generators. It takes a raw prompt, runs one‑click, multi‑round optimisation, and shows you before‑and‑after results together with evaluation metrics so you can see whether the revised prompt yields better model answers.

Core Capabilities (as described in the README)

Intelligent, iterative optimisation – one‑click runs several refinement rounds for both system‑level and user‑level prompts.
Evaluation suite – analysis, single‑result scoring, and multi‑result comparison to quantify improvement.
Multi‑model support – works with OpenAI, Gemini, DeepSeek, Grok, Zhipu AI, SiliconFlow, MiniMax and any OpenAI‑compatible custom endpoint.
Image generation mode – can optimise and run Text‑to‑Image, Image‑to‑Image and multi‑image generation using Gemini, Seedream, Grok, etc.
Prompt sources – start from scratch, templates, local files, or import from the community‑run Prompt Garden.
Smart Favorites – saved prompts keep version history, media, and source metadata; they can be exported/imported.
Advanced testing – variable substitution, multi‑turn conversation simulation, and function‑calling (tool use) testing.
Secure, client‑side architecture – the UI runs in the browser (or desktop app) and talks directly to the chosen model APIs; no intermediate server stores your data.
Multi‑platform delivery – available as a hosted web app, a downloadable desktop app (auto‑updating installers), a Chrome extension, or a Docker image. A password can protect deployments.
MCP (Model Context Protocol) server – exposes optimisation functions via a standard API, enabling integration with tools like Claude Desktop.

How to Get Started

Online demo – visit https://prompt.always200.com, add a model key in Model Manager, then follow the quick‑start tutorial.
Self‑host – one‑click Vercel or Cloudflare deployment, or run the Docker image (linshen/prompt-optimizer). Environment variables (VITE_OPENAI_API_KEY, ACCESS_PASSWORD, etc.) configure API keys and optional access control.
Desktop – download the installer (.exe, .dmg, .AppImage) from the GitHub releases page for a native app that avoids CORS limits.
Chrome extension – install from the Chrome Web Store and open the optimizer from the toolbar.

Typical Workflow

Create or import a prompt (text, template, or from Prompt Garden).
Run “Optimize” – the system sends the prompt to the selected LLM, receives a refined version, and repeats for a configurable number of rounds.
Evaluate – view analysis, compare scores across original and optimized outputs, and optionally run multi‑result tests.
Save – add the refined prompt to Favorites for reuse; export if you need to share.
Advanced testing – define variables, simulate multi‑turn dialogs, or test function‑calling before finalising.

Who Might Use It

Prompt engineers building reusable prompt libraries.
Developers integrating LLMs who need to fine‑tune prompts quickly.
Creators of text‑to‑image prompts who want more precise visual results.
Teams that want a self‑hosted, password‑protected UI for internal LLM usage.

Where to Find More

Docs – https://docs.always200.com (quick‑start, image‑mode, MCP server, deployment guides).
Source – https://github.com/linshenkx/prompt-optimizer (AGPL‑3.0 license, Docker image linshen/prompt-optimizer).
Community – Prompt Garden (https://garden.always200.com) for shared prompt assets.

TL;DR

Prompt Optimizer is a full‑stack, multi‑platform UI that lets you iteratively improve LLM or image‑generation prompts, evaluate the changes, and keep a library of refined prompts. It works with many major model providers, can be run locally or in the cloud, and exposes its optimisation functions via the Model Context Protocol for integration with other AI tools.