How it works

1. Capture, without ceremony

Open the app — on iPhone, iPad, Apple Watch, or Mac — and tap the capture orb. Say what's on your mind. Voice or video, ten seconds or ten minutes. There's nothing to title, tag, or file in the moment; the point is to get the thought out of your head before it goes.

On the watch it's one tap from your wrist, no phone required at capture time. On the Mac you can also record a window while you talk over it. Everywhere, capture works offline: thoughts queue on the device and upload when you're back on a network.

2. AI files it

Once a capture uploads, transcription turns your voice into text, and analysis turns the text into a structured note: a title, a summary, and any tasks, questions, or decisions it found inside. Your exact words are kept verbatim — the structure sits alongside them, never in place of them.

Then the thought is routed. ThoughtBridge knows your projects and picks the right one; when it isn't confident, the thought waits in a review inbox instead of guessing. You stay the editor — promoting a thought into a project is always a deliberate act, not something that happens to you.

3. Your agents continue

Filed thoughts are written into destinations — places you choose and control: a GitHub repo, Google Drive, or a local or iCloud folder. Notes arrive as plain text, Markdown, or JSON, and each destination carries agent files (AGENTS.md and friends) that tell AI tools what they're looking at.

That's the bridge: the next time you open Claude, ChatGPT, or Cursor in that project, your thinking is already there — dated, structured, and in context. One destination can host many projects, each in its own folder, so a single repo or Drive can hold your whole thinking surface.

And when you're ready to build: Bridge-it

Bridge-it closes the loop between a thought and shipped code. End a spoken thought with the words "bridge it", and a self-hosted runner — open source at launch — picks it up, drives Claude Code inside the project's repo, and opens a pull request.

If the agent has questions, they flow back to the app and you answer from wherever you are. The thought completes when the pull request merges. Nothing touches your code unless you've set up the runner yourself, on your own machine, for that project.

What stays yours

• Your words: captures are shown verbatim, never paraphrased away.
• Your files: destinations are repos, Drives, and folders you own.
• Your audio: recordings are transcribed, then the originals are deleted after a short retention window.
• Your call: nothing is promoted, published, or coded without a deliberate step from you.