Fluent

Product Introduction

1. Definition: Fluent is a native macOS AI writing assistant leveraging Modular Contextual Platform (MCP) and a proprietary Retrieval-Augmented Generation (RAG) engine. It operates as a system-level productivity tool integrating with any macOS application via its Smart Panel interface.
2. Core Value Proposition: Fluent eliminates app-switching chaos by enabling context-aware AI content generation directly within users’ active workflows. Its RAG engine replicates unique writing styles (personal, literary, or marketing voices) with 99% accuracy, eradicating generic "AI slop" while supporting 500+ AI models.

Main Features

1. Smart Panel:

   • How it works: A system-wide overlay activated via keyboard shortcut. Captures real-time context (selected text, clipboard, or active window content) using macOS Accessibility APIs.
   • Technology: Native Swift integration with macOS 14.2+, MCP for contextual awareness across 9+ built-in apps (Finder, Calendar) and 7,000+ external services.

2. Style Replication Engine (RAG):

   • How it works: Local RAG architecture indexes user-provided writing samples. During generation, it cross-references this "Memory" database via vector embeddings to mirror syntax, tone, and lexical patterns.
   • Technology: On-device vector storage (unencrypted), transformer-based similarity matching, and dynamic prompt injection.

3. Multi-Model Orchestration:

   • How it works: Users configure API keys (OpenAI, Anthropic, Mistral) or local models (Ollama, MLX). Actions route queries to specified providers via dynamic variable templating.
   • Technology: Unified API gateway supporting REST/WebSocket protocols, Apple Silicon-optimized MLX for local inference.

4. Document Intelligence:

   • How it works: Parses uploaded PDFs/images via OCR and semantic chunking. Enables Q&A, summarization, or style transfer from document content.
   • Technology: PDFTextExtractor framework, VisionKit for image OCR, and chunking via sliding-window tokenization.

5. Privacy-First Architecture:

   • How it works: All data processed locally unless using cloud APIs. History AES-256 encrypted; no telemetry or cloud processing.
   • Technology: macOS Keychain for API keys, CoreData with encryption for history, isolated sandbox for local models.

Problems Solved

1. Pain Point: Fragmented AI workflows requiring constant app switching, generic AI outputs lacking brand/personal voice consistency, and privacy concerns with cloud-based tools.
2. Target Audience:
   • Marketing teams needing on-brand copy across campaigns
   • Authors/journalists maintaining stylistic integrity
   • Privacy-conscious users (developers, legal professionals)
   • macOS power users seeking workflow automation
3. Use Cases:
   • Generating sales copy in a company’s tone directly from a CRM
   • Rewriting academic text in a researcher’s voice
   • Analyzing confidential contracts via local models
   • Batch-processing image-based reports into summaries

Unique Advantages

1. Differentiation:
   • Versus cloud tools (Jasper, Copy.ai): Full offline capability, no subscriptions, style replication.
   • Versus native apps (MacGPT): MCP context awareness, RAG personalization, and document intelligence.
2. Key Innovation:
   • MCP Integration: Unifies app-specific context (e.g., Calendar events, Reminders) into AI prompts dynamically.
   • Deterministic Styling: RAG engine achieves near-perfect style mimicry by bypassing fine-tuning via retrieval-based augmentation.

Frequently Asked Questions (FAQ)

1. Can Fluent run completely offline?
   Yes, Fluent supports 100% offline operation using local MLX models or self-hosted Ollama/LM Studio instances, with no data leaving your device.

2. How accurate is the writing style replication?
   Fluent’s RAG engine achieves ~99% style accuracy by cross-referencing user-provided writing samples during generation, preserving vocabulary, syntax, and tonal nuances.

3. What’s the cost for API-based model usage?
   Typical usage (editing/rewriting) costs $0.50–$1.00/month with OpenAI. Users control budgets via individual provider API keys and usage caps.

4. Does Fluent work with scanned documents or images?
   Yes, its document intelligence feature extracts text from PDFs and images via OCR, enabling semantic analysis and content generation from visual inputs.

5. How does licensing for teams work?
   Team licenses start at $29/machine (3+ Macs) with volume discounts. All plans include lifetime upgrades and cover commercial use across marketing/development teams.