PromptAmp — AI Prompt Enhancer 作者: Sina
Turn rough drafts into engineered prompts with one tap, in any text field. Bring your own API key — no backend, no telemetry.
扩展元数据
屏幕截图
关于此扩展
PromptAmp rewrites your rough draft into a well-engineered prompt, right inside the text field you are already typing in — on any website.
Type a rough idea. Press the button that appears in the corner of the field. PromptAmp shows you a better version next to your original, and replaces your text only when you accept it.
BRING YOUR OWN KEY
PromptAmp has no backend. Your draft goes directly from your browser to the AI provider you chose, authenticated with the API key you supplied. There is no PromptAmp server in between, because there is no PromptAmp server at all.
Works with OpenAI, Anthropic, Google Gemini, Groq, OpenRouter, local models via Ollama or LM Studio, and any endpoint that speaks the OpenAI chat-completions format — including your own self-hosted server or LiteLLM proxy.
Add several connections and order them: if the first is rate-limited or out of credit, the next takes over automatically and tells you it switched.
YOUR DRAFT IS NEVER TOUCHED UNTIL YOU ACCEPT
Not on error, not on timeout, not if the model declines. If insertion fails part-way through, the field is restored from a snapshot taken before anything started. Undo works twice over — the browser's own Ctrl+Z, plus an Undo button for ten seconds after.
WORKS IN REAL EDITORS
Not a whitelist of chat apps. Plain inputs, contenteditable, Monaco, CodeMirror, ProseMirror, Quill, Slate, fields inside iframes, fields inside modal dialogs. Five insertion strategies, each verified by reading the text back afterwards.
PROMPTS THAT FIT THE TASK
An image prompt is not a code prompt is not a study prompt. Seven built-in profiles — general, chat, image, video, coding, learning, writing — chosen automatically from the site you are on, or pinned by you. Fork any of them and edit the system prompt yourself.
Write your draft in any language. Optionally have the enhanced prompt come back in a different one, with your code, quoted strings, and generator parameters left byte-exact.
PRIVATE BY CONSTRUCTION
Open source under the MIT licence. Every claim above is checkable in the code: github.com/Sina-Amare/promptamp
COSTS
PromptAmp is free. You pay your own provider for the tokens you use — usually a fraction of a cent per enhancement. A daily cap is on by default so nothing can run away with your credit.
Type a rough idea. Press the button that appears in the corner of the field. PromptAmp shows you a better version next to your original, and replaces your text only when you accept it.
BRING YOUR OWN KEY
PromptAmp has no backend. Your draft goes directly from your browser to the AI provider you chose, authenticated with the API key you supplied. There is no PromptAmp server in between, because there is no PromptAmp server at all.
Works with OpenAI, Anthropic, Google Gemini, Groq, OpenRouter, local models via Ollama or LM Studio, and any endpoint that speaks the OpenAI chat-completions format — including your own self-hosted server or LiteLLM proxy.
Add several connections and order them: if the first is rate-limited or out of credit, the next takes over automatically and tells you it switched.
YOUR DRAFT IS NEVER TOUCHED UNTIL YOU ACCEPT
Not on error, not on timeout, not if the model declines. If insertion fails part-way through, the field is restored from a snapshot taken before anything started. Undo works twice over — the browser's own Ctrl+Z, plus an Undo button for ten seconds after.
WORKS IN REAL EDITORS
Not a whitelist of chat apps. Plain inputs, contenteditable, Monaco, CodeMirror, ProseMirror, Quill, Slate, fields inside iframes, fields inside modal dialogs. Five insertion strategies, each verified by reading the text back afterwards.
PROMPTS THAT FIT THE TASK
An image prompt is not a code prompt is not a study prompt. Seven built-in profiles — general, chat, image, video, coding, learning, writing — chosen automatically from the site you are on, or pinned by you. Fork any of them and edit the system prompt yourself.
Write your draft in any language. Optionally have the enhanced prompt come back in a different one, with your code, quoted strings, and generator parameters left byte-exact.
PRIVATE BY CONSTRUCTION
- No servers, no telemetry, no analytics, no accounts — ever
- Your API key is stored on this device and readable only by the background worker, never by the script that runs on web pages
- Keys are never synced through your browser vendor
- History stays local and can be cleared or turned off
- Host permissions are a short, fixed list of provider API hosts
Open source under the MIT licence. Every claim above is checkable in the code: github.com/Sina-Amare/promptamp
COSTS
PromptAmp is free. You pay your own provider for the tokens you use — usually a fraction of a cent per enhancement. A daily cap is on by default so nothing can run away with your credit.
评分 0(1 位用户)
权限与数据
必要权限:
- 访问您在所有网站的数据
可选权限:
- 访问您在 api.openai.com 的数据
- 访问您在 api.anthropic.com 的数据
- 访问您在 api.groq.com 的数据
- 访问您在 openrouter.ai 的数据
- 访问您在 generativelanguage.googleapis.com 的数据
- 访问您在 integrate.api.nvidia.com 的数据
- 访问您在 localhost:11434 的数据
- 访问您在 127.0.0.1:11434 的数据
- 访问您在 localhost:1234 的数据
- 访问您在 127.0.0.1:1234 的数据
根据开发者所述,必要的数据收集:
- 身份验证信息
- 网站内容
更多信息