# Plaud Transcripts Corrected With Your Own Terminology Glossary > Fixes the words your transcription keeps getting wrong, by giving the pipeline your vocabulary instead of hoping a bigger model guesses right. Every speech model mangles terms it has never seen: cell line and reagent names, drug and device names, case and matter numbers, part numbers, local spelling and number conventions, team and client names. Swapping to a different model does not fix this, because none of them have your terms either. Keep a plain Google Sheet glossary of the terms that matter, each with the spelling you want and the misfires it usually comes back as ("HEK293" for "heck two ninety three"), plus any formatting rules you need enforced, such as Swiss German ss instead of eszett or 1'000.00 number style. On each run the workflow reads the glossary, pulls the transcript of every new Plaud recording, applies the corrections to the transcript first, and only then writes the summary and action items from the corrected text, so your terminology is right everywhere downstream instead of only in the raw transcript. The corrected transcript and summary land in a Google Doc, and every correction made is listed so you can see what changed and keep tuning the glossary. Unknown terms are flagged, never silently normalized into something that looks plausible. Content type: workflow Source URL: https://www.agentpmt.com/agent-workflow-skills/plaud-transcripts-corrected-with-your-own-terminology-glossary Markdown URL: https://www.agentpmt.com/api/agent/workflows/plaud-transcripts-corrected-with-your-own-terminology-glossary?format=agent-md Updated: 2026-07-26T00:01:37.863Z Author: firef1ie --- Estimated time saved: 20 minutes. ## Tools - Get Users Current Time / Date - Google Sheets - Plaud - Google Docs Connector ## Workflow Outline 1. Get User Timezone and Date: Get the user's current local date, time, timezone and UTC offset. The requested window and every recording timestamp are resolved against this and never against the server clock. 2. Ask for the Recording Window: Goal: Establish the exact window of recordings this run covers, in the user's own timezone. | Inputs: The user's local date, time, timezone and UTC offset from the previous step, plus the last processed timestamp in the... 3. Read Terminology Glossary: Read the glossary tab: each row carries the correct term, the misheard variants it usually comes back as, and optional notes. Also read any formatting rules rows (spelling convent... 4. List Plaud Recordings: Call list_files with the widened UTC date_from and date_to from the window step. Plaud's timestamps are inconsistent and must be handled deliberately: the name field carries the r... 5. Each New Recording: Repeat this section for each recording. Correct one recording at a time. The glossary is applied to the transcript before anything is summarized from it. 6. Fetch Transcript: Call get_transcript with this recording's file id, reusing Plaud's existing transcript when one exists. 7. Summarize the Run: Goal: Report which recordings were corrected and which terms the glossary is still missing. | Inputs: The per-recording correction results, and the window that was used. | Outputs: The window covered stated in local tim... 8. Apply Glossary to Transcript: Goal: Correct the transcript against the user's glossary and formatting rules before any summary is produced. | Inputs: The raw transcript for one recording, and the glossary rows plus formatting rules read from the she... 9. Summarize From Corrected Text: Goal: Produce the summary and action items from the corrected transcript so the user's terminology is right downstream, not just in the raw text. | Inputs: The corrected transcript from the previous step, and the record... 10. Write Corrected Doc: Create a document for this recording, headed with its local recorded date and time, containing the summary and action items, then the corrected transcript, then a corrections tabl... 11. Log the Run: Append the recording id, its local recorded time, the doc link, the number of corrections applied, and any flagged unknown terms to the processed log so the recording is never han... ## Frequently Asked Questions ### How do I connect this workflow to my local agent like OpenClaw or Claude Code? You can install the local MCP server by opening a terminal and running: ``` npm install -g @agentpmt/mcp-router agentpmt-setup ``` This will connect you to local agents like Claude Code, Windsurf, Grok Build, Cursor, etc. Alternatively you can connect to the hosted version with this config block, no installation required: ``` { "mcpServers": { "agentpmt": { "type": "streamable-http", "url": "https://api.agentpmt.com/mcp", "headers": { "Authorization": "Bearer ", "x-instance-metadata": "{\"client\":\"generic-mcp\",\"platform\":\"remote\"}" } } } } ``` [View MCP Connection Instructions](/docs/mcp-reference/connection) for more details. ### How do I trigger this workflow? You can use it here in the browser by clicking the 'Try It' button. If you 'star' the workflow it will save to your dashboard for quick access. To use the workflow in an external agent, make sure you have followed the steps to connect it to the MCP server and then copy / paste this prompt, or just tell it to use the AgentPMT MCP server and use this workflow. > Call the AgentPMT-Workflow-Skills tool with action 'get\_workflow\_skill' and skill\_id 6a653502b640fe17224ac769 ("Plaud Transcripts Corrected With Your Own Terminology Glossary").