Plaud now ships both an official CLI and an official MCP server that cover almost everything this project does. New users should start there. This repo stays publicly available, but is in maintenance mode (see below).
Official CLI (docs):
npm install -g @plaud-ai/cli # requires Node.js >= 20Commands:
plaud login/logout/me,plaud files/recent/today/search,plaud file <id>,plaud audio <id>,plaud transcript <id> -o,plaud summary <id> -o.Official MCP server (docs) for Claude Code, Claude Desktop, Cursor, VS Code, Zed, ChatGPT, etc.:
npx -y @plaud-ai/mcp@latest install⚠ Binary name collision. Plaud's official CLI installs a binary called
plaud, the same name this project uses. Do not install both on the samePATH. If you are migrating,rm(or rename) this project's binary first.
This project Official replacement plaud loginplaud login(official)plaud logoutplaud logout(official)plaud listplaud files(orrecent,today,search)plaud download <id>plaud transcript <id> -o,plaud summary <id> -o,plaud audio <id>plaud syncNo official equivalent yet. The official CLI deliberately omits bulk operations: no
sync, no--watch, no prune, no on-disk archive layout, andsearchis capped at the most recent ~500 recordings. Audio is served via 24-hour presigned URLs, so building a long-lived offline archive against the official CLI means scripting your own downloader on top of it.If you need a scheduled, file-on-disk archive of every recording (audio + transcript + summary) in a stable
YYYY/MM/<slug>/layout (e.g. for GDPR data portability, compliance retention, or feeding a non-MCP pipeline),plaud syncis still the simplest path. For everything else, switch to the official tools.
- The repo stays public and the binaries stay downloadable.
- No new features. v0.3.0 acceptance will land; no new specs will be opened.
- Security fixes and Plaud-API breakage fixes will be considered on a best-effort basis.
- Issues and PRs may be triaged slowly. If your problem is something the official CLI or MCP can solve, please use those instead.
plaud-cli is NOT affiliated with, endorsed by, sponsored by, or connected to PLAUD LLC.
"PLAUD" and "Plaud.ai" are trademarks of their respective owners. This project uses these names solely to identify the third-party service it interoperates with. No claim of ownership, partnership, or endorsement is made or implied.
See
NOTICEfor the full statement.
A small, single-binary CLI for archiving recordings, transcripts, and summaries from your Plaud.ai account into local storage.
| Version | Scope | State |
|---|---|---|
| v0.1.0 | login, list, logout |
Released (2026-05-01) |
| v0.2.0 | download (transcripts, summaries, audio per recording) |
Released |
| v0.3.0 | sync (mirror your whole account, watch mode, prune) |
Implemented in main; pending acceptance walk |
| Project | Further development | Maintenance mode. Superseded by the official Plaud CLI and MCP server. |
Releases: https://github.com/simensollie/plaud-cli/releases
The roadmap and active specs live in specs/. No new specs will be opened.
Plaud.ai used to store your recordings, transcripts, and summaries in their cloud only. There was no sanctioned way to bulk-export or maintain a local archive. This tool filled that gap by talking to the Plaud web/mobile API with your own credentials and writing your data to disk in a format you control.
That gap is now largely closed. Plaud's official MCP server exposes recordings, transcripts, AI-notes, and presigned audio URLs to any MCP-compatible client. If you want conversational access to your Plaud data from Claude Code, Cursor, ChatGPT, etc., use that.
This tool remains useful only if you need a file-on-disk archive of every recording (audio + transcript + summary) in a deterministic directory layout, refreshed on a schedule, without a running AI client. Plaud's MCP does not currently offer that.
Two ways to install. Pick the one that fits.
plaud is a single static binary. No runtime dependencies, no installer, no admin rights needed. Pick your platform below; replace v0.2.0 with the tag from the Releases page if a newer one is available.
macOS (Apple Silicon):
curl -fsSL -o plaud.tar.gz \
https://github.com/simensollie/plaud-cli/releases/download/v0.2.0/plaud-cli_0.2.0_darwin_arm64.tar.gz
tar -xzf plaud.tar.gz plaud
mkdir -p ~/.local/bin && mv plaud ~/.local/bin/
plaud --versionFor Intel Macs, swap darwin_arm64 for darwin_amd64. If ~/.local/bin is not on your PATH, add it to ~/.zshrc or ~/.bashrc.
Linux (x86_64):
curl -fsSL -o plaud.tar.gz \
https://github.com/simensollie/plaud-cli/releases/download/v0.2.0/plaud-cli_0.2.0_linux_amd64.tar.gz
tar -xzf plaud.tar.gz plaud
mkdir -p ~/.local/bin && mv plaud ~/.local/bin/
plaud --versionFor ARM (Raspberry Pi etc.), swap linux_amd64 for linux_arm64.
Windows (PowerShell):
$tag = "v0.2.0"
$arch = "amd64" # or "arm64"
$url = "https://github.com/simensollie/plaud-cli/releases/download/$tag/plaud-cli_0.2.0_windows_$arch.zip"
Invoke-WebRequest -Uri $url -OutFile plaud.zip
Expand-Archive plaud.zip -DestinationPath .
.\plaud.exe --versionMove plaud.exe somewhere on your PATH. The simplest spot for a per-user install is %LOCALAPPDATA%\Programs\plaud\.
Verify the download (optional but recommended):
curl -fsSL -O https://github.com/simensollie/plaud-cli/releases/download/v0.2.0/checksums.txt
sha256sum -c checksums.txt --ignore-missingOK next to your archive's filename means the bytes match what the release pipeline produced.
You need Go 1.23 or newer.
git clone https://github.com/simensollie/plaud-cli
cd plaud-cli
go build -o plaud ./cmd/plaud
./plaud --versionA source build prints plaud version 0.x.y-dev (the development string) instead of the release version. Functionally identical.
For local development, the same toolchain runs the test suite:
go test -race ./...
go vet ./...
gofmt -l .See CLAUDE.md for contributor conventions and the spec-driven workflow.
Once plaud --version works, log in and try a few commands:
plaud login # interactive: region (us|eu|jp), email, OTP code
plaud list # show all recordings on your account, newest first
plaud download <id> # fetch one recording into ~/PlaudArchive (text by default)
plaud sync # mirror every recording on the account
plaud sync --watch --interval 15m # poll every 15 minutes (foreground; one terminal session)
plaud logout # delete stored credentialsFor Apple/Google SSO accounts (no password set), use the token-paste path:
plaud login --token <jwt> --region eu --email you@example.comThe full walkthrough — including how to grab the JWT from web.plaud.ai's localStorage — is in docs/user/getting-started.md.
| Item | Path |
|---|---|
| Credentials | ${XDG_CONFIG_HOME:-~/.config}/plaud/credentials.json (POSIX, mode 0600) %APPDATA%\plaud\credentials.json (Windows) |
| Archive root | ~/PlaudArchive/YYYY/MM/<slug>/ (POSIX) %USERPROFILE%\PlaudArchive\... (Windows) |
| Sync state file | <archive_root>/.plaud-sync.state |
Override the archive root with --out DIR per invocation, or PLAUD_ARCHIVE_DIR in your shell rc.
docs/user/— install, getting started, command reference, troubleshooting.docs/technical/— architecture, the (undocumented) Plaud API knowledge, contributor pointers.docs/README.md— the docs index.
This repo is spec-driven, test-driven, and "fail fast, fail often". See CLAUDE.md for principles, conventions, and the development workflow. See specs/README.md for how to read or write a spec.
In short: no code without a spec, every functional requirement has a test that cites its FR ID, tests come before implementation, and documentation in docs/ ships in the same change as the code.
MIT.
Use this tool only with your own Plaud.ai account. Using it against another person's account without authorization is unlawful. Running this tool against your own account is data portability under GDPR Article 20.