Plaud MCP Integration
Plaud MCP Tools — Quick Start
You have access to 11 MCP tools, 4 prompts, and 12 skill resources for managing Plaud AI recordings, transcripts, and memory. Each tool returns structured JSON with next_steps guidance.
Prerequisites
- MCP server configured in
.mcp.json— authenticated via GitHub OAuth (automatic) ELEVENLABS_API_KEYenv var — only needed for ElevenLabs transcription
Tool Quick Reference
| Tool | Purpose | Skill |
|---|---|---|
find_recordings |
Browse and filter recordings | recording-search |
get_recording |
Full recording metadata | recording-search |
get_audio_url |
Presigned download URL (5min expiry) | recording-search |
transcribe |
ElevenLabs transcript (cached or live) | transcription |
get_content |
AI summary or meeting notes | transcription |
trigger_processing |
Plaud server-side AI processing | transcription |
get_account_info |
Account profile and membership | – |
get_processing_status |
AI processing queue status | – |
list_languages |
Supported language codes | transcription |
memory_search |
Semantic search across transcripts | memory-search |
memory_ingest |
Bulk index recordings into memory | bulk-operations |
Prompts
MCP clients can discover and invoke these workflow templates:
| Prompt | Purpose | Parameters |
|---|---|---|
daily_briefing |
Status overview — recent recordings, queue, memory | None |
analyze_transcript |
Deep analysis of a recording | file_ref |
search_memory |
Guided dual-path search | query, time_range (default “7d”) |
bulk_summary |
Summarize multiple recordings | count (default 5) |
File ID Resolution
All tools accepting file_ref support three formats:
| Format | Example |
|---|---|
| Full ID | 4f757af256ecba4fab502739c122dc78 |
| Short prefix | 4f75 |
| Row number | 3 (from last find_recordings) |
Row numbers persist across tool calls within a session — call find_recordings once, then use row numbers in subsequent transcribe, get_content, etc. calls.
Cost Safety
Tools that consume external credits prompt for confirmation via MCP elicitation:
transcribe— asks before calling ElevenLabs API (cached transcripts are free)memory_ingest(source="all")— asks before bulk operations that may trigger transcriptions
Older MCP clients that don’t support elicitation skip the prompt and proceed directly.
Skills as Resources
The MCP server exposes plugin skills as MCP resources — any client can discover and read them without installing the plugin:
skill://recording-search/SKILL.md
skill://memory-search/SKILL.md
skill://transcription/SKILL.md
skill://bulk-operations/SKILL.md
skill://using-plaud-mcp/SKILL.md
skill://using-plaud-cli/SKILL.md
Setup
The MCP server is configured in the plugin’s .mcp.json. Authentication uses GitHub OAuth — Claude handles the flow automatically on first connect. See the plugin README for details.
See Also
- recording-search — Browsing, filtering, and metadata retrieval
- memory-search — Semantic search across indexed transcripts
- transcription — ElevenLabs transcription with cost awareness
- bulk-operations — Batch ingestion, auto-polling, and export
- using-plaud-cli — CLI usage guide (
plaudcommand)