StenoAI / Enterprise
Private intelligence.
Institutional control.
Give your teams the clarity of AI meeting notes while keeping control of their conversations. On the device or on infrastructure you control, open source, and never on ours.
02 / Where it’s used
A private record. A shared purpose.Built for the conversations
that carry consequence.
When information is sensitive, the tools in the room matter. Give your teams useful AI without handing over their conversations.
Government
For the public interest.Within your perimeter.
Policy discussions, ministerial briefings and committee meetings. Keep the record where it belongs, with the people responsible for it.
Defence
Clarity in the room.Continuity in the field.
Planning sessions, programme reviews and debriefs. Capture decisions on your own hardware, even when a connection isn’t an option.
Law enforcement
Accountable records.Inside the force.
Investigation reviews, command meetings and witness contact notes. A transcript behind every summary, kept under your own retention rules.
03 / Where it runs
One pipeline. Three boundaries.Your device, or your
infrastructure. Never ours.
Choose where the AI runs to match the sensitivity of the conversation. StenoAI operates no server that receives your meeting content, in any mode.
On the device
Capture, transcription, summaries and questions all run on the laptop. After setup it works offline and on air-gapped networks.
On your infrastructure
Point summaries at a model server on your own network, such as Ollama or any OpenAI-compatible endpoint you host. Transcription stays on the device unless you also point it at your own speech endpoint.
In your cloud tenancy
Use your organisation’s own cloud account, such as AWS Bedrock or your OpenAI or Anthropic agreement, or route through your organisation’s gateway so desktops never hold provider keys.
Open-weight models on the device
- Gemma 4 E2B 2B
- Gemma 4 E4B 4B
- Llama 3.2 3B
- Qwen 3.5 9B
- Gemma 4 12B
- GPT-OSS 20B
Downloaded once during setup, then they run offline. Or point it at your own endpoint.
04 / Verify, don’t trust
Don’t take our word for it.
Check it.
Every claim on this page can be checked against the source. Every step happens on your device, or on infrastructure you control, and the files stay within the boundaries you set.
Inspect the sourceRead the code that touches the audio.
MIT-licensed source. Your security team can review every line and every dependency.
Trust a build, not a binary.
Compile it in your own environment and evaluate it against your own requirements.
Pull the cable. It keeps working.
Once the app and models are installed, local mode runs with no network connection at all.
We tell you what does go out.
The full list is below. Meeting content only ever goes to an endpoint you run or choose.
Capture
Mic + system audio
Process
Your AI models
Understand
Notes + answers
Retain
Your files
What leaves the device
- Meeting content
- Never in local mode. Otherwise only to the model endpoint you configure.
- Update checks
- Version check only. No meeting content.
- Usage analytics
- Anonymous and switchable off in Settings. No meeting content.
- Model downloads
- During setup, or when you choose a new model.
06 / Get started
Two editions. One pipeline.From one laptop to the
whole organisation.
Try the open-source app on a single laptop today. When a department, unit or force is ready, we take you from briefing to deployment.
Community
Free · MIT licence
- The full on-device pipeline
- No account and no usage caps
- macOS · Windows alpha
Enterprise
For teams and organisations
- 01Brief
A briefing for your team and security lead.
- 02Pilot
A pilot on your own meetings and hardware.
- 03Assure
Source, build and data paths for your security review and DPO.
- 04Deploy
Roll out on devices or your infrastructure, through your procurement route.
Your next conversation deserves better.
Keep the record.
Keep control.
See what private meeting intelligence
can do for your organisation.