TranscriptionTools

TranscriptionTools is an MCP server that enables AI clients like Claude Desktop to format, correct, and summarize raw speech-to-text transcripts. Designed for researchers, podcasters, writers, and software developers working with transcribed spoken audio, it provides automated tools to turn messy automated transcripts into clean, readable text. The server operates directly on raw strings or file paths on the local filesystem, without requiring access to the original audio files. Through its exposed tools, TranscriptionTools strips awkward timestamps and introduces paragraph breaks and line gaps based on the natural duration of pauses in conversation. It also features contextual text repair, evaluating speech transcriptions against semantic context to spot homophones or transcription errors and correct them with a ninety percent confidence threshold. In addition, users can request structured summaries constrained by word counts, character limits, or calculated speaking durations based on estimated speaking rates. Repair logs are preserved so agents and users can review previous changes across sessions. By linking LLMs directly to file-based or inline transcript data, it streamlines post-transcription editing and synthesis workflows.

Category: Design, Media & Creative

Tags: audio, summarization, text-processing, transcription

Visit TranscriptionTools

How to install and configure TranscriptionTools

1. Install the server for Claude Desktop via Smithery: bash npx -y @smithery/cli install @MushroomFleet/transcriptiontools-mcp --client claude 2. Alternatively, clone the repository and build manually: bash git clone https://github.com/mushroomfleet/TranscriptionTools-MCP cd TranscriptionTools-MCP npm install npm run build 3. Add the server configuration to your MCP settings file: json { "mcpServers": { "transcription-tools": { "command": "node", "args": ["/path/to/TranscriptionTools-MCP/build/index.js"], "disabled": false, "autoApprove": [] } } } 4. Update /path/to/TranscriptionTools-MCP/build/index.js with your absolute file path and restart your MCP client.

What you can do with TranscriptionTools

• Formatting timestamped transcripts: Strip raw timestamps from audio transcription text files while converting timing pauses into natural paragraph and line spacing. - Correcting transcription errors: Run contextual repairs on misheard words and grammatical slips without original audio, using high-confidence semantic analysis. - Auditing transcript revisions: Retrieve detailed repair logs using session identifiers to review every semantic substitution and spelling correction made by the tool. - Generating length-constrained summaries: Produce concise summaries of lengthy interview or meeting transcripts tailored to specific word counts, character limits, or speaking durations.

Key facts

• https://github.com/MushroomFleet/TranscriptionTools-MCP
• Design, Media & Creative, Files, Documents & PDFs
• audio, summarization, text-processing, transcription

Part of MCP Servers

What can TranscriptionTools do?

TranscriptionTools processes transcripts through four primary functions: converting timestamped transcripts into natural paragraphs based on pause length, repairing transcription errors using semantic context, retrieving historical repair logs by session ID, and generating summaries constrained by word counts, character lengths, or speaking durations.

How do I install TranscriptionTools?

You can install TranscriptionTools automatically using the Smithery CLI command with Claude Desktop. Alternatively, clone the GitHub repository, install dependencies using npm, compile the TypeScript source files with npm run build, and register the node executable and build path inside your client MCP settings JSON file.

Which MCP clients work with TranscriptionTools?

TranscriptionTools works with any client supporting the Model Context Protocol, such as Claude Desktop. Once registered in the client settings file with the Node.js execution command pointing to its compiled entry point, its four tools become available directly in agent conversations.

Does TranscriptionTools require original audio files?

No, TranscriptionTools operates purely on text input or text file paths. It analyzes semantic context to identify errors, strips timestamps directly from the text, and calculates pause gaps using recorded timestamp markers without needing access to the original recordings.

Is TranscriptionTools open source?

Yes, TranscriptionTools is open source software released under the MIT license. The full TypeScript source code, configuration files, and documentation are publicly available on GitHub, allowing developers to inspect the implementation, customize formatting thresholds, or contribute modifications.