Privacy-aware AI workspace

Your workspace, your models.See what your AI is doing.

Review sensitive details before they leave your browser. Get answers that cite your files. Choose your models and see what each reply cost.

Explore the sample See it in action

A fictional walkthrough. Opens in a new tab.

A real SecureMind review: the outgoing message contains [email redacted], with options to edit or send the reviewed text. — Real interface. Fictional example.Watch the review

Open source and self-hosted

Run it yourself. It is open source.

SecureMind is an open-source generative AI security tool. You can run it yourself today: get the code from GitHub, start it, and connect the models you already use, whether that is your own API keys, Amazon Bedrock, or local models. A desktop app and a hosted option are planned.

On your own computer

Pull the code from GitHub and start it with Docker. Your workspaces, files and memory stay on that computer. The only traffic that leaves is to the model providers you choose to connect. An API key is strongly recommended, because the built-in local models are slow on most machines. A desktop app for Windows and Mac is planned.

In your own cloud

If you would rather host it for your team, the same code deploys to your own AWS account with Terraform. Sign-in is invite-only, the data lives in your account with backups and alerts, and it suits teams of 1 to 25 people, or up to about 100 with a raised AWS limit. Turn on Amazon Bedrock, or add an API key, so chat has a model.

Not available

We host it for you

A separate AWS account for you, on your own domain and with your own name and logo, set up and kept up to date for you. Not available. The Install page has the details and a way to say you are interested.

Open source

The whole project is on GitHub under the Apache 2.0 license, so you can read exactly what it does before you run it.

Free and open source (Apache 2.0). SecureMind is just the tool: you add your own API keys, Amazon Bedrock access or local models, and pay those providers directly. It is a personal project shared as is, so review what you send and check that it meets your own requirements.

View the code on GitHub How to run it

Watch it work

The full walkthrough

About 23 minutes, from workspaces to settings, with chapters so you can jump to what you care about. All data shown is made up.

Nothing loads from YouTube until you press play.

Watch on YouTube

Jump to a chapter

Inside SecureMind

A closer look

Explore one feature at a time. Choose a view below, then take a closer look at how it works.

Product recording · fictional demo data01 / 05

01 / Before you send

Check what leaves your browser

If you type something that looks sensitive, like an email address or a phone number, SecureMind stops and shows you what it found. You see what will be sent in its place before anything goes out.

Detection runs in your browser, before the message is sent.
You see the replacement text and approve it.
You can edit the message or send the redacted version.

This lowers the risk but it can miss things, so give the review a quick look.

Open the sample site in a new tab

Set it up your way

Your models, your name, your team

A quick tour of the screens you use to set it up.

Product recording · fictional demo data01 / 04

01 / Connect your models

Add a key and see what it unlocks

Paste an API key from Anthropic, OpenAI, Gemini, Groq or Grok and SecureMind recognises the vendor, checks the models that key can use, and lists the ones that work. Amazon Bedrock and local Ollama models sit in the same list.

Paste a key and the vendor is recognised straight away.
The key is checked by testing the models it lists.
Saved keys show only their last few characters.

Keys are stored encrypted on your own install. Bedrock uses your AWS sign-in, so no key is pasted for it.

Open the sample site in a new tab

Take the next step

See how it fits your work.

Explore the sample with fictional data, or run your own personal workspace.

View sample Run it yourself