Prompt refinement modes
Polysh ships with multiple styles of refinement: highly detailed, casual, precise and direct, creative boost, structured, instructional, role-based, few-shot, zero-shot, chain of thought, and custom mode.
Client-Side Prompt Refinement Engine for Browser Environments
Polysh improves prompts before they reach the model. It lives inside the browser, works across major AI chat sites, and gives users precise control over tone, structure, and intent without sending data through a middleman.
Polysh ships with multiple styles of refinement: highly detailed, casual, precise and direct, creative boost, structured, instructional, role-based, few-shot, zero-shot, chain of thought, and custom mode.
Before a prompt is injected, Polysh can surface a diff view so the user can inspect what changed and choose whether to accept or discard the refined text.
Everything stays client-side. Credentials, preferences, prompt history, and usage stats are stored locally in browser storage and dispatched directly to the selected provider endpoint.
A dedicated shortcut lets users invoke refinement instantly from the keyboard, which keeps the experience quick for power users and frequent writers.
Download the ZIP file, extract it, and load the unpacked extension from your Chromium-based browser.
Download ZIPUse the Mozilla Add-ons listing for the easiest install, or use the signed XPI for direct installation.
Open Firefox Add-onsThe extension is built from the repository source, with all assets and configuration files stored in the project root.
View sourcePolysh helps users turn raw ideas into clearer instructions, reusable templates, and more deterministic outputs. That means fewer retries, cleaner workflows, and better results across the tools people already use every day.