SaidTrace

Product Introduction

1. Definition: SaidTrace is a macOS application (Mac app) that functions as a local-first audio archive and AI-powered note-taking system. It is technically categorized as a personal knowledge management (PKM) tool and a local transcription client for macOS.
2. Core Value Proposition: SaidTrace exists to give users complete ownership and privacy over their recorded audio and derived data. Its core proposition is enabling users to record on their preferred device (iPhone, Apple Watch, digital recorder) and process the audio locally on their Mac using their choice of AI model (local or external) for transcription, summarization, and Q&A, all while keeping the original recordings and data on their personal hardware.

Main Features

1. Local-First Audio Archive: SaidTrace creates a centralized searchable archive on your Mac. It imports audio via AirDrop (from iPhone), direct connection (from recorders), or local file import. The app uses OpenAI's Whisper model to perform local transcription entirely on-device, ensuring audio never leaves your computer. Transcripts, notes, and a search index are stored locally in the archive.
2. AI-Agnostic Note & Ask Engine: The app decouples audio processing from AI analysis. After local transcription, you can choose which AI processes the text. It supports local AI models via LM Studio or Ollama for private processing. For external AI services (like ChatGPT, Claude), SaidTrace sends only the necessary text context for your query, not the audio file. The "Ask" feature allows you to question your recordings, with answers traceable back to the source audio timestamp.
3. Model Context Protocol (MCP) Integration: SaidTrace provides an MCP server, allowing AI coding assistants and tools (like Cursor, Claude Code, Codex CLI) to search your local recording archive directly. The MCP connection returns relevant text snippets and recording IDs to the AI client, enabling it to reference your past conversations or notes within its workflow, without ever accessing the raw audio bytes.

Problems Solved

1. Pain Point: It solves the privacy and vendor lock-in problem of cloud-based voice note apps. Many services upload audio to their servers, use proprietary AI, and charge per minute, creating ongoing costs and data security concerns.
2. Target Audience: The primary user personas are privacy-conscious professionals (journalists, researchers, therapists, lawyers), knowledge workers who conduct frequent interviews or meetings, and developers/technologists who prefer local-first, open-ecosystem tools and want to integrate their notes into AI-assisted coding workflows via MCP.
3. Use Cases: Essential for transcribing confidential client interviews locally, creating searchable archives of lecture or meeting recordings without a cloud subscription, and enabling AI pair programmers to reference past verbal project discussions or stand-up meeting notes directly within their IDE.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros: Unmatched privacy through on-device Whisper transcription; flexibility to use any AI model (local or cloud); one-time purchase model with transparent $3/month or $18/year pricing after free tier; no per-minute fees; integrates with existing Apple ecosystem (iPhone, iCloud for sync) and AI developer tools via MCP.
   • Cons: Requires a macOS 14+ Apple Silicon Mac for local processing, locking out Windows/Linux and Intel Mac users. The workflow requires manual import steps (like AirDrop) rather than seamless automatic sync from iOS. Advanced features depend on user setup of third-party tools (LM Studio/Ollama for local AI, Node.js for MCP).

2. Key Alternatives & Differentiation:

   • Omi / Plaud Note / Fieldy: These are hardware+subscription services. They require purchasing a dedicated device ($129-$179) and paying a high monthly fee ($20-$30) for cloud processing. SaidTrace differentiates by using hardware you already own (iPhone, recorder) and charging a significantly lower software-only fee ($3/month) for the archive and AI orchestration layer.
   • Descript / Otter.ai: These are cloud-native, subscription-based editors. They upload all audio to their servers, use their own AI, and focus on collaborative editing. SaidTrace differentiates by being a private, local archive first, with AI as a modular feature, and no collaboration focus.
   • MacWhisper / other local transcribers: These are one-time purchase apps focused solely on local Whisper transcription. SaidTrace differentiates by adding the structured archive, the "Ask" feature with multi-AI support, and the MCP server integration for developer workflows.

Frequently Asked Questions (FAQ)

1. Does SaidTrace upload my audio to the cloud? No. SaidTrace is a local-first application. The OpenAI Whisper model transcribes audio directly on your Mac. Your audio files, transcripts, notes, and search index never leave your computer. Only if you choose to use an external AI service (like OpenAI's API) is the necessary text from a query sent to that service.
2. What AI models can I use with SaidTrace? You have multiple choices. For complete privacy, you can use a local large language model (LLM) via LM Studio or Ollama installed on your Mac. Alternatively, you can connect to external AI APIs. SaidTrace acts as the orchestrator, sending prompts and context to your model of choice.
3. Can I use SaidTrace on my iPhone or Windows PC? The core SaidTrace archive application is currently only available for macOS 14+ on Apple Silicon Macs. You record on your iPhone or other device and import the files to the Mac app. There is no standalone iOS or Windows application at launch.
4. How does the MCP (Model Context Protocol) integration work? SaidTrace runs a local MCP server on your Mac. AI tools that support MCP (like Cursor IDE) can connect to it. This allows those tools to search the titles and notes of your SaidTrace archive. When you ask your AI a question, it can use this tool to find relevant past recordings and reference them in its answer.
5. What's included in the free monthly allowance? The free plan includes 20 audio imports and 20 AI summaries/Ask queries per month. This allows you to test the full workflow with your own recordings. Paid plans ($3/month or $18/year) remove these limits.