# SpeakPen

> SpeakPen is a voice note app. You record a thought, it is transcribed with AssemblyAI and given a title and summary, and the note is written in the language you spoke. Notes can be read in the app, in the web app, and in an Obsidian vault through the SpeakPen Sync plugin. It is available on iPhone, on Android, and in the browser.

## Product

- Platforms: iPhone app (App Store), Android app (Google Play), web app at https://speakpen.app/app, and the Obsidian plugin "SpeakPen Sync" (Obsidian community plugin store).
- Each recording is transcribed and gets a title and a summary. The note is in the language that was spoken.
- Obsidian: SpeakPen Sync brings your SpeakPen notes into an Obsidian vault as Markdown. It fetches only notes changed since the last sync. A note edited in SpeakPen is rewritten in place and keeps the filename it was created with. A note you edited yourself in the vault is not overwritten. The plugin only reads from SpeakPen and never pushes vault changes back.
- Landing page languages: English, Simplified Chinese, Traditional Chinese, Japanese, Korean, Spanish, French, German, Portuguese, and Italian.

## Pricing

- Free: 3 recordings per month, up to 5 minutes each.
- Pro: $9.99 per month, 60 recordings per month, up to 20 minutes each.
- Prices are as shown on the landing page (https://speakpen.app/#pricing).

## MCP server

- Endpoint: https://speakpen.app/mcp (Streamable HTTP, stateless).
- Authentication: OAuth 2.0 (scope `notes:read`), or a SpeakPen API token as a Bearer token.
- Access is read-only. Tools: `search`, `fetch`, and `list_recent_notes`. They search, fetch and list the signed-in user's own notes. The server cannot create, change or delete notes, and it never receives audio.
- Rate limit: 60 tool calls per minute per account.
- MCP registry name: io.github.Qumge/speakpen-mcp
- Setup guide for every client: https://speakpen.app/connect (plain Markdown: https://speakpen.app/connect.md).

## Paid audio and video capabilities through Qumge

These standalone tasks accept new audio or a public YouTube URL without a SpeakPen account. Connect to https://qumge.com/mcp using the buyer's Qumge authentication. Each operation includes every internal processing step; Qumge charges the fixed price only after complete delivery. The account MCP above continues to read existing SpeakPen notes only.

| Capability | Operation | Input limit | Fixed USD price per complete result |
| --- | --- | --- | --- |
| [SpeakPen Structured Transcription](https://qumge.com/en/caps/speakpen-structured-transcription) | `transcribe_audio` | 30 minutes, 100 MiB | $0.60 |
| [SpeakPen Meeting Minutes](https://qumge.com/en/caps/speakpen-meeting-minutes) | `create_meeting_minutes` | 30 minutes, 100 MiB | $0.65 |
| [SpeakPen Audio Subtitles](https://qumge.com/en/caps/speakpen-audio-subtitles) | `create_subtitles` | 30 minutes, 100 MiB | $0.55 |
| [SpeakPen Audio to Notes](https://qumge.com/en/caps/speakpen-audio-to-notes) | `create_audio_note` | 20 minutes, 100 MiB | $0.49 |
| [SpeakPen YouTube Overview](https://qumge.com/en/caps/speakpen-youtube-transcript) | `get_youtube_overview` | Token-budget preflight; short public YouTube video | $0.10 |

- YouTube overview: one public HTTPS `video_url`, optional output `language` (`zh` default or `en`). Gemini 3.8 Flash processes the video on the server; returns topic, summary, 3–6 points and limitations. No transcript or timestamp guarantee. Inaccessible or over-budget videos fail; no partial-video success. You must have permission to process the supplied video.
- Audio input: one directly downloadable HTTPS `audio_url`. A signed URL works; keep it available for at least one hour. Page URLs, playlists, live streams and cookies are unsupported.
- Results: full timestamped transcription; meeting minutes with grounded actions and decisions; complete SRT and WebVTT file contents; or a Markdown voice note with its full transcript. Output stays in the source language.
- Call `get_cap` with the selected slug to read its complete input/output schemas and current terms before purchase.
- Call `call_cap` with the slug, operation, input and a unique `idempotency_key` for this order. Reuse that key only when retrying the same operation and input.
- An accepted or running job is not a delivered result. Read its platform job ID with `get_cap_job`; status and result reads are free. Save returned subtitle `content` using its `filename`, or save the returned `markdown` directly.
- Provider manifest: [qumge.json](https://speakpen.app/.well-known/qumge.json).

Example `call_cap` arguments:

```json
{
  "slug": "speakpen-structured-transcription",
  "operation": "transcribe_audio",
  "input": { "audio_url": "https://example.com/recording.m4a" },
  "idempotency_key": "a-unique-key-for-this-recording-order"
}
```

Then call `get_cap_job` with `{ "job_id": "ac_..." }`, using the job ID returned by Qumge.

## Docs and code

- [speakpen-mcp](https://github.com/Qumge/speakpen-mcp): Public documentation for the SpeakPen MCP server.
- [speakpen-skill](https://github.com/Qumge/speakpen-skill): Agent skill for working with SpeakPen notes.

## Pages

- [SpeakPen home (English)](https://speakpen.app): Product overview and pricing.
- [Connect SpeakPen to Obsidian and AI agents](https://speakpen.app/connect): Step-by-step setup for Claude Code, Codex, Cursor, VS Code, Gemini CLI, ChatGPT, Claude and Obsidian.
- [Connect (Markdown)](https://speakpen.app/connect.md): The same guide as plain Markdown, for agents.
- [SpeakPen home (Simplified Chinese)](https://speakpen.app/zh-Hans): Landing page in Simplified Chinese.
- [SpeakPen home (Traditional Chinese)](https://speakpen.app/zh-Hant): Landing page in Traditional Chinese.
- [SpeakPen home (Japanese)](https://speakpen.app/ja): Landing page in Japanese.
- [SpeakPen home (Korean)](https://speakpen.app/ko): Landing page in Korean.
- [SpeakPen home (Spanish)](https://speakpen.app/es): Landing page in Spanish.
- [SpeakPen home (French)](https://speakpen.app/fr): Landing page in French.
- [SpeakPen home (German)](https://speakpen.app/de): Landing page in German.
- [SpeakPen home (Portuguese)](https://speakpen.app/pt): Landing page in Portuguese.
- [SpeakPen home (Italian)](https://speakpen.app/it): Landing page in Italian.
- [Voice to Obsidian](https://speakpen.app/voice-to-obsidian): Guide page.
- [Voice to Obsidian (German)](https://speakpen.app/de/voice-to-obsidian): Translated guide page.
- [Voice to Obsidian (Simplified Chinese)](https://speakpen.app/zh-Hans/voice-to-obsidian): Translated guide page.
- [Voice to Obsidian (Korean)](https://speakpen.app/ko/voice-to-obsidian): Translated guide page.
- [Voice to Obsidian (Japanese)](https://speakpen.app/ja/voice-to-obsidian): Translated guide page.
- [AudioPen Alternative](https://speakpen.app/audiopen-alternative): Guide page.
- [Obsidian Transcription](https://speakpen.app/obsidian-transcription-plugin): Guide page.
- [Voice Memos to Obsidian](https://speakpen.app/iphone-voice-memos-to-obsidian): Guide page.
- [AudioPen to Obsidian](https://speakpen.app/audiopen-obsidian-sync): Guide page.
- [Obsidian Quick Capture](https://speakpen.app/obsidian-quick-capture): Guide page.
- [Siri Recap & Obsidian](https://speakpen.app/siri-recap-obsidian): Guide page.
- [Plaud Alternative](https://speakpen.app/plaud-alternative): Guide page.
- [Obsidian Dictation on Mac](https://speakpen.app/obsidian-voice-dictation-mac): Guide page.

## Blog

Articles on capturing ideas by voice and turning them into output with an AI agent. Index: https://speakpen.app/blog (Simplified Chinese: https://speakpen.app/zh-Hans/blog).
- [Why Your Second Brain Became a Notes Graveyard (and How to Fix It)](https://speakpen.app/blog/why-your-second-brain-became-a-notes-graveyard): Capturing notes got cheap; processing them never did. Why second brains turn into graveyards, and four ways to make your notes useful again.
- [Five Moments Your Best Ideas Get Lost (and How to Keep Them)](https://speakpen.app/blog/where-good-ideas-get-lost): Good ideas rarely show up at your desk. Here are the five moments knowledge workers lose them most often, and what to say into your phone so they survive.
- [From Voice Note to Finished Draft: A Workflow With Your AI Agent](https://speakpen.app/blog/turn-voice-notes-into-output-with-ai-agent): Record ideas by voice in SpeakPen, connect your AI agent to your notes, and use three copy-paste prompts to turn them into plans and drafts.
- [How to Turn Voice Memos Into Blog Posts (That Still Sound Like You)](https://speakpen.app/blog/turn-voice-memos-into-blog-posts): Talking a draft is faster than typing one. Here's a four-step method to turn voice memos into blog posts without AI sanding off your voice.
- [Stop Re-Explaining Yourself to AI: Let Your Voice Notes Be the Context](https://speakpen.app/blog/stop-re-explaining-yourself-to-ai): Every new AI chat starts from zero. Build a context library from voice notes you own and can read, and let your agent pull just what it needs.
- [Recorded Dozens of Voice Memos You Never Listen To? How to Clear the Backlog](https://speakpen.app/blog/clear-your-voice-memo-backlog): Voice memos pile up because audio can't be skimmed. How to deal with the old backlog honestly, and record new ideas so reviewing takes minutes.
- [Claude Code + Obsidian as a Second Brain: Where Voice Notes Fit](https://speakpen.app/blog/claude-code-obsidian-second-brain-voice-notes): Obsidian stores, Claude Code reasons, but your vault only holds what you typed at a desk. Here's the capture layer that gets your walks and commutes in.
- [Capture First, Think Later: How to Record Ideas So Your Agent Can Use Them](https://speakpen.app/blog/capture-first-think-later): Separate capturing ideas from thinking about them. Seven habits for recording ideas by voice so you and your AI agent can actually use them later.

## Account and legal

- [Privacy Policy](https://speakpen.app/privacy)
- [Terms of Service](https://speakpen.app/terms)
- [Delete your account](https://speakpen.app/delete-account): How to delete a SpeakPen account.