The platform
Platform
What Deep Thought Cloud does with everything you give it: every source processed together into one knowledge base, and where you and your AI agents use it.
Every source, one knowledge base
A recording, a meeting or a file is a fragment. However it comes in, it takes the same path, and the platform processes it against everything already known. Transcription is where the platform starts, not where it ends.
SourcesWhere it comes in
- record-me (iOS) In beta
- Trillian (macOS) In alpha
- File upload: images, PDFs and documents Premium
- Recording in the browser
- Microsoft TeamsIn development
- File upload
- Folder sync (Dropbox) (on request) Enterprise
The platformEvery source processed together
Deep Thought Cloud
- Takes in Transcription of audio and video · Speaker diarization
- ACI — Adaptive Conversational Intelligence
- Understand what each source is, and how sensitive
- Connect people, organisations and topics, tied to your world
- Extract decisions, commitments, questions, quotes
Over time
A living picture of each person, project and topic, current as each new source arrives.
- People
- Projects
- Topics
- Decisions
KnowledgeWhere you and your AI agents use it
- Web App
- MCP server Premium
- REST API Enterprise
By stage
Sources
The Web App takes file uploads and records in the browser. record-me records on the iPhone (Android in testing), Trillian brings in voice memos from the Mac, and the REST API accepts uploads from your own systems.
The meeting notetaker (in beta; Premium and Enterprise) joins a Zoom, Google Meet or Teams meeting and records its audio.
Accepted: audio and video.
Folder sync from Dropbox is set up by the operator, on request. It is not self-serve.
Transcript, speakers, title
Every source is transcribed and its speakers are separated, on every plan. Both run on own hardware in Stockholm.
Every source also gets a title from a local model. This is the whole of the Free plan: the transcript, who spoke, and a title.
Knowledge: ACI
With Premium, Adaptive Conversational Intelligence works through each source in three levels: what the source is, who and what it is about, and what in it matters. Across all sources, it keeps a picture of each person, project and topic that every new source updates. That picture, not any single transcript, is what the platform builds.
Where each step runs, including AI analysis, is stated on the processing page. Processing in Sweden
Where you read it

Web App
Upload, record in the browser, read transcripts, search and share.
MCP server
Connect an AI assistant to your knowledge — one server, OAuth sign-in, no command line.
REST API
The same platform the apps use, for your own systems.
Your notes
Trillian writes each transcript into a folder of plain text files on your Mac, next to your own notes. Free.
What it does not do
- Free does not include ACI. It stops at the transcript, the speakers and the title.
- Folder sync is not self-serve. The operator configures it per account.
- Trillian runs on macOS only. There is no Trillian for Windows or the web.
- An extraction is never detached from its source: each one points back to where it came from.
The whole chain: five steps
The platform is the middle of a longer chain. Each tool does one thing in it.
1 Said
A conversation is captured where it happens.
record-me (iOS), Trillian (macOS), File upload: images, PDFs and documents, Recording in the browser, File upload, Folder sync (Dropbox) (on request)
2 Understood
It becomes text, speakers and knowledge, every source processed together.
Deep Thought Cloud: Transcription of audio and video, Speaker diarization, ACI
3 Asked
You and your AI agents ask what is known, and get answers that point back to the source.
Web App, MCP server, REST API
4 Kept
Your own copy, in your own files: Trillian brings each transcript back into your Obsidian vault on your Mac, and core-skills, a separate family of Claude Code skills, distils it into tasks and insights.
Trillian, core-skills
5 Surfaced
Marvin reads what is kept and hands you the one thing you missed. A reply you prepare there goes back out, to a chat or a mail, where it is said again.
Steps 4 and 5 run on your own Mac. They are not part of Deep Thought Cloud and need no plan.
See it in practice: ten use cases, each with the app that does the work and the plan it needs. Use cases