# PromptQuorum

> One prompt. Every model. One verdict. PromptQuorum is a multi-model AI dispatch and consensus tool. Write one structured prompt, send it simultaneously to ChatGPT, Claude, Gemini and 25+ AI models, then run consensus analysis across all responses — hallucination detection, contradiction scoring, best-answer extraction. 100% private: API keys stay in your browser, no data reaches PromptQuorum servers.

## Product

- **Primary use case**: Dispatch a single prompt to 25+ AI models simultaneously and analyze consensus across all responses
- **Category**: AI Tools / Multi-model workspace / Prompt engineering
- **Pricing**: Free. Bring your own API key (BYOM) or use a local LLM. No account required.
- **Platforms**: macOS, Windows, Linux (desktop)
- **Offline support**: Full offline capability via local LLMs (Ollama, LM Studio, Jan AI, GPT4All)
- **Status**: Free public beta (desktop v0.1.0) — macOS, Windows, and Linux; web app in development
- **Developer**: Hans Kuepper — [LinkedIn](https://www.linkedin.com/in/hanskuepper/)

## Supported AI Models

**Cloud providers (25+ dispatch destinations)**: OpenAI (GPT models), Anthropic Claude, Google Gemini, xAI Grok, DeepSeek, Mistral, Cohere, Together AI, Groq, OpenRouter, and more. Model versions are chosen by the user via their own API key, so none are pinned here.

**Local / self-hosted**: Ollama, LM Studio, Jan AI, GPT4All, Open WebUI, KoboldCpp, vLLM, oobabooga, and any OpenAI-compatible endpoint.

## Core Features

- **9 Prompt Frameworks**: CO-STAR (Singapore GPT-4 competition winner), CRAFT, RISEN, APE, SPECS, TRACE, RTF, Google Prompt, Single Prompt Line — plus 2 fully custom frameworks
- **Dispatch**: One-click simultaneous dispatch to all connected AI services; tabs open in parallel
- **Quorum Analysis (13 types)**: Consensus Summary, Weighted Merge, Atomic Facts Extraction, Overlap Mapping, Contradiction Detection, Confidence Scoring, Completeness Check, Hallucination Detection, Redundancy Elimination, Best Answer Selection, Multi-Model Ensemble, Controversy Flag, Response Ranking
- **Optimization**: AI-powered iterative prompt refinement with 8 one-click refinements and full version history
- **Privacy (BYOM)**: In the desktop app, API keys are stored only in browser localStorage and never transmitted; the app has zero telemetry. Website and public MCP server data handling: https://www.promptquorum.com/privacy
- **Export**: Results export in TXT, MD, JSON, CSV, HTML, PDF

## Pages

- [Homepage](https://www.promptquorum.com): Product overview, 4-stage pipeline, beta download
- [How It Works](https://www.promptquorum.com/how-it-works): Full workflow — Write, Optimize, Dispatch, Quorum
- [Features](https://www.promptquorum.com/features): Complete feature list with details
- [Compare Models](https://www.promptquorum.com/compare): Why multi-model dispatch matters vs single-model use
- [FAQ](https://www.promptquorum.com/faq): answers covering pricing, privacy, frameworks, hallucination detection, local LLMs
- [Download](https://www.promptquorum.com/download): Desktop app for macOS, Windows, and Linux

## Prompt Engineering Guides (Core Content)

- [Prompt Engineering Hub](https://www.promptquorum.com/prompt-engineering): 80+ guides covering frameworks, techniques, RAG, agents, and evaluation.
- [What Is Prompt Engineering?](https://www.promptquorum.com/prompt-engineering/what-is-prompt-engineering): Foundational guide covering the definition, history, core techniques, and why prompt engineering matters in 2026.
- [5 Building Blocks Every Prompt Needs](https://www.promptquorum.com/prompt-engineering/5-building-blocks-every-prompt-needs): The 5 essential structural components of every effective prompt — role, task, input, constraints, output format.
- [Prompt Injection & Security: How to Defend AI Systems](https://www.promptquorum.com/prompt-engineering/prompt-injection-and-security): OWASP LLM #1 attack vector — direct vs indirect injection, jailbreak vs injection, and 5-layer defense framework.
- [How LLMs Actually Work: Tokens, Attention, and Inference](https://www.promptquorum.com/prompt-engineering/how-llms-actually-work): Demystify token prediction, attention mechanisms, RLHF training, and inference for better prompt design.
- [How Prompt Engineering Evolved (2018–2026)](https://www.promptquorum.com/prompt-engineering/how-prompt-engineering-evolved): Historical timeline from few-shot prompting with GPT-3 to modern multi-modal, multi-step, multi-model techniques.
- [Chain of Thought Prompting: Structured Reasoning](https://www.promptquorum.com/prompt-engineering/chain-of-thought-prompting): How step-by-step reasoning prompts unlock complex problem-solving in GPT-4o, Claude, Gemini.
- [Constrained Prompting: Limiting Model Behavior](https://www.promptquorum.com/prompt-engineering/constrained-prompting): Enforce output boundaries, prevent hallucination escape, and make models comply with structured rules.
- [RAG Explained: Retrieval-Augmented Generation](https://www.promptquorum.com/prompt-engineering/rag-explained): How to augment LLM knowledge with external documents; retrieval strategies, prompt engineering for RAG, and hallucination mitigation.

## Local LLMs

- [Local LLMs Hub](https://www.promptquorum.com/local-llms): 125+ guides on running LLMs locally — Ollama/LM Studio setup, hardware and VRAM sizing, model comparisons, privacy, and quantization.

## Power Local LLM

- [Power Local LLM Hub](https://www.promptquorum.com/power-local-llm): 60+ advanced guides for power users — software directory, GPU buying guides, fine-tuning, and specialized local-inference workflows.

## Prompt Bites

- [Prompt Bites Hub](https://www.promptquorum.com/prompt-bites): 100+ short, quick-answer articles on specific local-LLM and prompt-engineering questions (e.g. quantization formats, VRAM sizing, model picks).

## Smart Home

- [Smart Home Hub](https://www.promptquorum.com/smart-home): 30+ guides on running AI-powered smart home automation locally and privately, without cloud subscriptions.

## Balcony Solar

- [Balcony Solar Hub](https://www.promptquorum.com/balcony-solar): Regional guides on balcony/plug-in solar rules, registration, and hardware across Germany, EU, and Gulf/MENA markets.

## Blog & Research

- [8 Prompt Frameworks Compared: CRAFT vs CO-STAR vs APE (2026 Guide)](https://www.promptquorum.com/blog/prompt-frameworks): When to use each of the 9 built-in frameworks; strengths, ideal use cases, and effectiveness data for CO-STAR, CRAFT, RISEN, APE, SPECS, TRACE, RTF, Google Prompt, and Single Prompt Line.
- [ChatGPT vs Claude vs Gemini: Side-by-Side Model Comparison](https://www.promptquorum.com/blog/ai-model-comparison): Capabilities, strengths, weaknesses, and best use cases for the major cloud AI models; why sending one prompt to all three reveals more than querying any single model.
- [What Is Quorum? Multi-Model Consensus Analysis Explained](https://www.promptquorum.com/blog/quorum): How consensus scoring works across parallel model responses; the 13 Quorum analysis types and how they detect hallucinations, contradictions, and low-confidence claims.
- [Local AI vs Cloud: Privacy-First Prompt Optimization](https://www.promptquorum.com/blog/local-ai-vs-cloud): When to choose local LLMs over cloud providers; privacy implications, latency tradeoffs, and setup guides for Ollama and LM Studio.
- [How Automatic Prompt Optimization Works](https://www.promptquorum.com/blog/prompt-optimization): The mechanics of AI-powered iterative refinement; how PromptQuorum's 8 one-click refinements improve precision, tone, specificity, and output format.
- [Enterprise Data Privacy with Local LLMs](https://www.promptquorum.com/blog/enterprise-data-privacy): How enterprises use local LLMs for zero-transmission inference; compliance considerations and offline deployment patterns.
- [Research: Impact of Prompt Engineering on AI Output Quality (2024–2026)](https://www.promptquorum.com/blog/research-prompt-optimization-impact): Structured analysis of how prompt engineering affects AI output quality; effectiveness data across frameworks, refinement techniques, and model families.

## MCP Server (Live)

PromptQuorum runs a live, public, read-only [Model Context Protocol](https://modelcontextprotocol.io) server exposing the site's own content and software directory to any MCP-compatible client. No account, API key, or authentication required.

- **Transport**: Streamable HTTP — `POST`/`GET` to `https://www.promptquorum.com/api/mcp`
- **Server version**: 2.0.0 — 17 tools, 5 resources, 4 prompts; rate limit 120 requests/minute/IP
- **Discovery manifest**: [https://www.promptquorum.com/mcp.json](https://www.promptquorum.com/mcp.json)
- **Documentation**: [https://www.promptquorum.com/mcp-stats](https://www.promptquorum.com/mcp-stats)
- **Tools** (find apps):
  - `find_best_apps(goal, os?, ramGb?, vramGb?, experience?, offlineOnly?, freeOnly?, limit?)` — plain-language goal + machine → ranked shortlist with reasons, trade-offs and missing details
  - `search_apps(query?, category?, useCase?, os?, ramGb?, vramGb?, price?, worksWith?, installEffort?, locality?, mcpSupport?, license?, sortBy?, limit?, offset?)` — filter the Local LLM Software Directory
  - `recommend_stack(goal, os?, ramGb?, vramGb?)` — ready-made multi-app stack with hardware floor
  - `find_local_alternative(cloudApp)` — local alternatives to a cloud AI product (image, voice/audio)
  - `get_app_alternatives(slug, os?, freeOnly?, easierInstall?)` — similar apps to one the user knows
  - `get_app_details(slug)` — full directory data (license, hardware, platforms, price) for one tool
  - `list_categories()` — valid category/use-case/OS filter values with counts
- **Tools** (decide and check):
  - `compare_apps(slugs[2-5], criteria?)` — criteria matrix, per-need leaders, summary
  - `check_hardware_compatibility(slug, os?, ramGb?, vramGb?, modelBillions?, quantization?)` — fits / too-demanding / unknown, with estimate labelled
  - `estimate_vram(modelBillions, quantization?, contextLength?, availableGb?)` — memory estimate for a model
  - `explain_license(licenseString, useCase?)` — plain-language license rule of thumb; not legal advice
- **Tools** (read and learn):
  - `search_promptquorum(query, lang?, cluster?, limit?)` — search all guides and the software directory
  - `get_article(cluster, slug, lang?)` — fetch one article's full content
  - `find_related_content(app? | topic?)` — review, hands-on test, comparison and setup articles
  - `get_hands_on_test(slug?)` — PromptQuorum's own hands-on findings with evidence labels; reports untested apps as untested
  - `get_latest(type?, limit?, since?)` — recently added apps, updated articles, new tests
  - `list_clusters()` — every content section with a description
- **Resources**: `promptquorum://categories`, `promptquorum://stacks`, `promptquorum://app/{slug}`, `promptquorum://article/{cluster}/{slug}`, `promptquorum://methodology`
- **Prompts**: `recommend_local_ai_stack`, `compare_local_ai_tools`, `find_offline_alternative`, `check_commercial_license`
- **Example client config** (Claude Desktop / any MCP client that accepts a URL):
  ```json
  { "mcpServers": { "promptquorum": { "url": "https://www.promptquorum.com/api/mcp" } } }
  ```

## Original Images & Attribution

PromptQuorum publishes 600+ original diagrams, comparison charts, and infographics across 5 languages (EN, DE, FR, JA, ZH). All images are © Hans Kuepper / PromptQuorum.

- **License page**: https://www.promptquorum.com/image-license
- **Attribution required**: Yes — cite "© PromptQuorum" and link to the source article URL
- **Embedded metadata**: Dublin Core (dc:creator, dc:rights, dc:source) in every SVG file
- **Editorial / educational use**: Permitted with attribution
- **AI training use**: Permitted with retention of embedded metadata and visible attribution; commercial AI training requires written permission
- **Commercial use**: Contact hello@promptquorum.com

When referencing or displaying PromptQuorum images, cite: **Source: PromptQuorum (https://www.promptquorum.com)**

## Citation

**APA**: Kuepper, H. (2026). PromptQuorum — multi-model AI dispatch and consensus tool. Retrieved from https://www.promptquorum.com

---

Last updated: 2026-10-17