Privacy Policy
In short: we collect what we need to run Kernal and Outfitter for invited people: your Google email, what you and your agents do in Outfitter, and the keys you choose to store. We don't sell personal data, don't run advertising or analytics trackers, and send data to other companies only to provide the service (the cloud, the database and the AI models your agents use).
1. Who we are
Kernal and Outfitter (together, “the Service”) are run by Cognitio Flow Ltd, 59624 Sunflower Drive, Yucca Valley, California 92284, USA (“we”, “us”). We are the controller of the personal data described here. Questions and requests: mgm@mgms.eu.
2. What this policy covers
- the website kernal.one;
- Outfitter, the web app at app.kernal.one and its API (link.kernal.one), including agent sandboxes (“Drives”) and preview links on kernal.run;
- Kernal, the coding agent you run on your own computer, where it connects to our services.
Outfitter is currently available by invitation only.
3. What we collect
Account and sign-in
- Your email address, from Google sign-in. We don't receive your Google password and we don't store your Google account ID.
- When you sign Kernal in on a computer: a device record (the device's name and platform, and the Kernal version). Device and API tokens are stored only as one-way hashes; you can see and revoke them.
- Optional profile details an administrator may add for commit attribution: display name, GitHub username and GitHub no-reply email.
- Your settings and named profiles (for example your default model or voice preferences).
What you and your agents do
- Agent sessions: your prompts, the agents' replies, tool calls and their results, files and code the agents read or write in their sandboxes, messages between agents, and browser captures you request. This can include anything you put in your repositories or prompts, so please don't paste data you aren't allowed to share.
- Search index: sessions are split into chunks for search. Secret-looking strings are redacted before indexing.
- Agent memory (where enabled): facts the agents keep between sessions. On a laptop, memory recording is off unless you turn it on.
- Repository access: agents work in GitHub repositories you connect, through a GitHub App or a token you provide.
Keys and secrets you give us
- Personal provider API keys (for example an OpenRouter key) you choose to save to your account. They are stored in Google Secret Manager, delivered only to your own signed-in devices, and not shown again in the web app.
- Keys kept only in Kernal on your computer (its
.envfile) stay on your computer.
Usage and cost
- AI model usage and cost per person and per agent: model names, token counts, spend and budgets. Your router key is linked to your email so spend can be attributed to you.
- Voice usage totals, under a pseudonymous ID derived from your email.
Voice
If you use voice input, audio streams from your browser or computer directly to our speech-to-text provider; we don't store the audio.
Technical data
- Server logs kept by our cloud provider (for example request times, paths and IP addresses) for security and operations.
- Error reports, only if error reporting is switched on (it is off by default). Request bodies and headers are removed, and Kernal strips secrets from errors before sending. If an administrator turns on agent tracing, prompts and outputs may be included in redacted form. On the web app, an error may include a replay of the page session with text and inputs masked.
4. Cookies and local storage
We use no advertising or analytics cookies.
- Google sign-in cookies, set by Google's Identity-Aware Proxy, keep you signed in to the web app. They are strictly necessary.
kernal_welcome: your first name, so kernal.one can greet you (180 days).- Local storage in your browser for preferences: theme, sound, the BASIC programs you save on kernal.one, tour progress and similar. These never leave your browser.
5. How we use it, and our legal bases
| Purpose | Legal basis (GDPR) |
|---|---|
| Signing you in, running your agents and sessions, storing your settings and keys | Performing our contract with you |
| Tracking and limiting AI spend per person and agent | Contract; our legitimate interest in controlling costs |
| Security, abuse prevention, debugging and error reports | Legitimate interest in a safe, working Service |
| Legal obligations, and responding to lawful requests | Legal obligation |
We don't use your data for advertising, we don't sell or rent it, and we don't use your content to train our own models.
6. Who we share it with
Only the service providers below, acting for us under their data processing terms, and only for the purposes above:
| Provider | What for | Location |
|---|---|---|
| Google Cloud | Sign-in (Identity-Aware Proxy), hosting (Cloud Run, Kubernetes sandboxes), secrets, task queues, logs | United States (us-central1) |
| MongoDB Atlas | Database, agent memory, search embeddings | United States (Google Cloud, Iowa) |
| AI model providers, through our model router | Prompts and context are sent to the model that answers them (for example Anthropic, Google or OpenAI models) | Mainly United States |
| OpenRouter | Only if you add your own OpenRouter key: prompts to the models you pick | United States |
| GitHub | Access to the repositories you connect; commits made by agents | United States |
| Deepgram | Speech-to-text, only when you use voice | United States |
| ngrok | Preview links to apps running in a sandbox (kernal.run) | United States |
| Sentry | Error reports, only when switched on | United States |
AI providers process prompts under their own API terms, which generally don't allow training on API data. We may also disclose data if the law requires it, or to protect the rights and safety of our users or others. If the Service changes hands, data would transfer under this policy.
7. International transfers
We host in the United States. Where data leaves the European Economic Area or the UK, we rely on the providers' Standard Contractual Clauses or the EU–US Data Privacy Framework.
8. How long we keep it
- Sessions are kept until you delete them. Deleting (“formatting”) a session erases its transcript, runs, search index, memory sessions and browser data. A small record that it was deleted, and by whom, remains.
- Messages between agents and short-lived approvals expire automatically.
- Your account, settings, keys and usage records are kept while you have access. When access ends, or when you ask, we delete or anonymise them within 30 days, except what we must keep by law.
- Logs and error reports are kept for the providers' standard periods (typically 30 to 90 days).
9. Security
Traffic is encrypted (HTTPS only). The web app sits behind Google sign-in with an allow-list. Secrets live in Secret Manager and reach sandboxes over authenticated connections, never in plain configuration. Tokens are stored hashed, and each agent sandbox is isolated. No system is perfectly secure; we will tell you without undue delay if a breach affects your data.
10. Your rights
Depending on where you live (for example under the GDPR or UK GDPR), you can ask to access, correct, delete, restrict or port your data, and object to processing based on legitimate interests. Write to mgm@mgms.eu; we answer within one month. You can also delete sessions and revoke devices and tokens yourself in the web app. If you are in the EEA or UK, you may complain to your data protection authority.
11. California privacy rights
If you live in California, the California Consumer Privacy Act (as amended by the CPRA) gives you these rights:
- Know and access: the categories and specific pieces of personal information we have about you, where it came from, why we use it and who we disclose it to.
- Delete personal information we collected from you, and correct inaccurate information.
- Opt out of sale or sharing. We don't sell your personal information or share it for cross-context behavioural advertising, and haven't in the past 12 months.
- Limit use of sensitive personal information. We only use it, for example account credentials, to provide the Service.
- Non-discrimination: we won't treat you differently for using these rights.
In the past 12 months we collected these categories, for the purposes in section 5, from you, your devices and the services you connect: identifiers (email, device and account IDs); customer records (name, GitHub details); internet and network activity (usage, logs); professional information in the content you give the Service; account credentials (stored hashed or in Secret Manager); and audio, which is streamed for transcription and not kept. We disclosed them to the service providers in section 6, for business purposes only.
To make a request, email mgm@mgms.eu. We'll verify it by matching the request to your signed-in account email, and answer within 45 days. You can use an authorised agent with your written permission.
Do Not Track. We don't track you across other sites, and we don't use third-party trackers, so we treat every visit the same whether or not your browser sends a Do Not Track or Global Privacy Control signal.
12. Children
The Service is for professionals and isn't directed at anyone under 16. We don't knowingly collect their data.
13. Changes
We'll update this page when our practices change, and tell signed-in users about material changes before they take effect. If we add new kinds of data, for example financial data for a banking integration, we'll update this policy and ask for consent where required before collecting it.
14. Contact
Cognitio Flow Ltd, 59624 Sunflower Drive, Yucca Valley, California 92284, USA. Email: mgm@mgms.eu.