6sense
Ask an AI agent about 6sense account insights, buying signals and keyword trends.
Read Plaud voice-recorder transcripts, AI notes and audio.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The Plaud MCP server connects the recordings made by Plaud's hardware voice recorders — the Note, Note Pro, NotePin and NotePin S — to Claude, ChatGPT and any MCP-compatible agent. Anthropic's directory lists five tools and all five read. Sign-in is OAuth 2.0 with PKCE, and it reaches transcripts, AI notes and the audio itself.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
scopes_supported entirely. There is no permission the consent screen can show you and no boundary between reading a transcript and any future capability.get_file returns a presigned download URL valid 24 hours. A transcript can be redacted before it reaches an agent; a voice recording cannot.list_files call.tools/list response. The endpoint returned HTTP 401 to an anonymous request. The tool names, schemas and safety annotations here come from the published source of @plaud-ai/mcp 0.3.9 — strong evidence, and still not the running server's own reply.Not reliably, and Plaud says the duty is yours. Its help centre states you are responsible for informing all participants and obtaining any legally required consent, and that Plaud does not collect consent on your behalf. The NotePin S shows a red LED that cannot be turned off, but an LED on a worn pin is notice only if someone looks at it.
Transcripts, AI notes and the audio itself. The five tools return recording metadata, the timestamped transcript with speaker attribution, the AI summary with action items and key topics, and a temporary audio download URL. Plaud's help centre confirms the assistant gets a download link valid for 24 hours alongside the transcript and summary.
No. All five tools carry readOnlyHint true and destructiveHint false in the published server source. Plaud's own export skill states plainly that Plaud MCP does not expose a push tool, and its documentation lists what the connector cannot do: it cannot record audio and it cannot edit recordings or transcripts. There is no delete path.
None are declared. The server's RFC 9728 protected-resource descriptor and its authorization-server metadata were both read on 2026-08-23 and neither document carries a scopes_supported field. Nothing at the consent screen narrows the grant, so authorizing the connector authorizes the whole tool surface as one unit, now and as it grows.
Until you delete them or close the account. Plaud's security article states recordings, transcripts and summaries are retained until the user manually deletes them or cancels their subscription, with no time-based expiry. Cloud storage is described as unlimited, so a broad list_files call reaches every synced recording you have ever made.
Five on the remote server, and Plaud's own table lists seven. The two extras, login and logout, are registered only by the local stdio entry point in the published npm package. The hosted endpoint this page describes registers the same five the directory lists, so the two lists agree once the transports are separated.
A Plaud account with Private Cloud Sync switched on. Plaud states that MCP works by giving your assistant access to your synced recordings, so cloud sync must be enabled, and that Plaud MCP is currently free for all active accounts. Terminal installs additionally need Node.js 20 or later.
Yes, and that is the documented reason to use it. Plaud's help centre states its OAuth API deliberately returns only your 20 most recent recordings, and directs anyone needing full history to MCP or the CLI, neither of which is subject to that cap. MCP is the widest read path Plaud offers an individual user.
.md route). That host's robots.txt carries Content-Signal: ai-train=yes, search=yes, ai-input=yes — AI synthesis expressly permitted. · retrieved 2026-08-23llms.txt (retrieved 2026-08-23) · retrieved 2026-08-23@plaud-ai/mcp 0.3.9, published 2026-08-20 — registry metadata and tarball read from (the npmjs.com web page and its robots.txt both sit behind a Cloudflare challenge that returns 403 to our client, so the registry API was used instead). Tool names, descriptions, input schemas, safety annotations and the stdio/HTTP split were read from the package's own dist bundle and its bundled skills/ directory (retrieved 2026-08-23). · retrieved 2026-08-20help_center JSON API after the HTML returned 403 to our client) · retrieved 2026-08-23/pages/trust-center is a 301 to this URL) · retrieved 2026-08-23www-authenticate Bearer challenge naming https://mcp.plaud.ai/.well-known/oauth-protected-resource/mcp; that descriptor and the authorization server metadata at https://mcp.plaud.ai/.well-known/oauth-authorization-server were both read, and a nonsense control path under the same prefix returned 404, confirming no interpolating catch-all (2026-08-23) · retrieved 2026-08-23Connect Plaud once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.