History

9 snapshots

Tracking since Apr 9, 2026.

View as table

Date	Users	Rating	Reviews	Version
Apr 9, 2026	9	—	—	1.4.0
Apr 19, 2026	11	—	—	1.4.0
Apr 25, 2026	11	—	—	1.4.0
May 2, 2026	13	—	—	1.8.0
May 9, 2026	15	—	—	1.8.0
May 19, 2026	13	—	—	1.8.0
May 25, 2026	12	—	—	1.8.0
Jun 1, 2026	6	—	—	1.8.0
Jun 8, 2026	9	—	—	1.8.0
Now	8	—	—	1.8.0

Changelog

Apr 25, 2026

description

Prompt Temple Broadcaster is a research tool for systematically comparing 
AI model behavior, capabilities, and reasoning across multiple providers.

RESEARCH CAPABILITIES:
- Parallel prompt execution across different model architectures
- Response comparison for bias, accuracy, and reasoning quality
- Template-based reproducible testing across trials
- Local storage for data privacy during sensitive research
- Export and analysis of response patterns

TECHNICAL FEATURES:
- Works with ChatGPT, Claude, Gemini, Grok, and more
- Manifest V3 compliant extension
- All processing on user's machine; no data collection
- Open protocol for adding new AI providers
- Works offline with saved templates

RESEARCH APPLICATIONS:
→ Model behavior analysis
→ Prompt engineering optimization
→ Cross-model bias detection
→ Reasoning quality comparison
→ Template library building

Designed for researchers who need reproducible, systematic AI evaluation.

Stop testing AI models one at a time.

Prompt Temple Broadcaster sends your question to several major chat assistants at once so you can see how different models reason about the same problem, side by side, in seconds.

But first — get the prompt right.

A live in-page quality panel and a guided prompt builder work directly inside the assistant you are already using. You can tell whether your prompt is weak before you hit send, rewrite it in one click, or walk through a short question-and-answer flow to turn a rough idea into something production-ready. All of it without leaving the chat window.

Why researchers and prompt engineers use it

One prompt, many models. Run parallel tests across different model architectures without switching tabs or copy-pasting.

See the cracks. Expose bias, reasoning gaps, and accuracy failures that only appear when models are compared head to head.

Write prompts that don't waste credits. A live quality panel scores your prompt for clarity, specificity, and structure as you type. One click rewrites it. No more burning API calls on vague questions.

Ask before you prompt. AskMe is a guided four-step wizard that turns "write me a plan" into a fully structured, spec-complete prompt with quality metrics. Answer a handful of targeted questions and walk away with a prompt that reads like it was written by a professional.

A library at your fingertips. Search the Discover catalog of over 9,000 curated prompts semantically from the side panel. Copy, iterate, broadcast. Build on what already works.

Reproducible by design. Save prompt templates, run consistent trials, and build a library others can replicate. Everything runs locally by default, with optional cloud sync when you sign in.

Your data stays yours. Local-first with zero telemetry. Nothing leaves your machine unless you explicitly sign in for sync.

Export-ready results. Capture and analyze response patterns for papers, reports, or client work.

How it works

1. Write smarter prompts in place. Visit your favorite assistant. A subtle sigil appears next to the composer. Type your prompt and watch the clarity, specificity, and structure scores update in real time.

2. Enhance or refine. Press "Enhance" for a one-shot rewrite (3 credits) or "Refine with AskMe" to launch the guided wizard (2 + 3 credits) that produces a fully specced prompt with a quality report.

3. Broadcast everywhere. Once the prompt is ready, fire it at every model you want to compare. Open all provider tabs from the popup, authenticate once, broadcast, and watch responses stream in parallel.

4. Compare, save, iterate. Use Compare Wall to tile provider tabs into a grid so you can watch every response arrive side by side. Save anything useful to your library. Search the Discover catalog for prompts that have already solved similar problems.

Features at a glance

The Prompt Studio offers a live quality panel that scores every prompt you type on clarity, specificity, and structure; a one-click rewrite via the Prompt Temple engine; a guided builder that moves from intent to clarifying questions to synthesis, producing a production-ready prompt with a quality score, spec completeness, and improvement metrics; semantic search across more than 9,000 curated public prompts; and a personal library where you can save, tag, search, and sync prompts across devices.

The Broadcast Engine runs parallel broadcasts to multiple assistants in one click, tiles provider windows into a grid so you can watch every response stream side by side, pulls scattered provider tabs back into one window, restores yesterday's test run, and exports comparison results as JSON, Markdown, or CSV.

Multi-Engine Research lets you query several search engines at once and feed the results into your broadcast as context, with a radar-style UI that shows each engine responding as its results come back.

Supported assistants

The extension broadcasts to the major public chat assistants available today. The list expands over time as new providers add compatible composers. You can broadcast to as many or as few as you pick for a given run — authentication is handled once per provider in your browser.

Built for serious work

Manifest V3 compliant. Local-first architecture. No account required to use the broadcaster or the live quality panel — you only sign in if you want cloud library sync or enhancement credits. No telemetry. No data collection.

The extension is useful for anyone who needs to compare model behavior, benchmark reasoning quality, iterate quickly on prompts without losing their place, or build a shareable prompt library for a team.

If you have ever wondered whether a different model would answer your question better, now you can find out in one pass. And if you have ever wondered whether a prompt is even good enough to send, the quality panel will tell you before you hit enter.

Send one prompt to several assistants at once. Compare results side by side. A live quality panel helps you write better prompts.

Apr 25, 2026

short_description

Broadcast your prompts to 12 AI platforms simultaneously. Powered by Prompt Temple.

Broadcast your prompts to 13 AI platforms simultaneously. Powered by Prompt Temple.

Apr 25, 2026

host_permissions

https://api.prompt-temple.com/*

https://api.prompt-temple.com/*, https://export.arxiv.org/*, https://api.github.com/*, https://api.stackexchange.com/*

Apr 25, 2026

permissions

activeTab, tabs, storage, scripting, clipboardWrite, sidePanel, alarms, identity

activeTab, tabs, storage, scripting, clipboardWrite, sidePanel, alarms, identity, system.display, downloads

Permissions & access

Permissions: activeTabtabsstoragescriptingclipboardWritesidePanelalarmsidentitysystem.displaydownloads
Host access: https://api.prompt-temple.com/*, https://export.arxiv.org/*, https://api.github.com/*, https://api.stackexchange.com/*

Screenshots

About