Privacy Policy · Semantix

Effective 7 October 2026 · Version 1.2

Semantix is a sole proprietorship based in Israel. We decide what happens to the personal data described on this page, which makes us its controller. Write to privacy@semantixstudio.com about any of it.

The short version#

Your code lives on your computer. The editor, the code index, and the graph work fully offline and never upload it.

When you run an agent, work is sent. That request carries your prompt, the file context for it, the results of tools the agent runs, and your provider key — to our service, which passes it to the model provider you chose. Local only is on unless you turn it off, and while it is on we keep none of it.

What we keep beyond that, you switch on. Turn off Local only and we keep your agent runs. Turn on Back up conversations or Semantix Memory and we keep those too. Tasks keep their schedules and run history with us, because that is how they follow you.

Your account holds your name, your email, and a hashed password, or the Google or GitHub account you sign in with. If you pay, Paddle takes the payment; we never see your card.

The rest of this page is the detail behind those paragraphs. Section 3 of our Terms of Use says the same things, on purpose.

What goes where#

	Where it runs	What we keep
The editor, index and graph	Your computer	Nothing. It works offline.
Agent requests, with Local only on	Our service, then your model provider	Nothing. Used for that request only.
Agent runs, with Local only off	Our servers	Conversations, tool calls and their results, token usage
Back up conversations (off until you turn it on)	Our servers	A copy of each conversation
Semantix Memory (off until you turn it on)	Our servers	Memories, turn captures, session notes
Tasks	Our servers and your computer	Schedules, run history, the end of each run's output
This website and your account	Our servers	Name, email, hashed password or sign-in provider, plan
Payments	Paddle	Paddle holds your payment details. We keep your plan and subscription status.

1. What never leaves your computer#

Your source code and project files. The Studio reads and indexes the files you point it at. That index is written to a directory on your own computer. It is not uploaded and we have no access to it.
The code graph. Built locally, from the local index.
Your settings, themes, and local configuration.
Your API keys at rest. They are written to a configuration file in your own home directory with owner-only permissions.

The editor and the graph require nothing from us. If our servers vanished tomorrow, they would keep working.

2. What is sent when you run an agent#

Running an agent is a request to send work to a model. That is what it is for, and it cannot be done without sending something. Here is exactly what.

Sent with each agent request:

Your prompt.
The file context assembled for it, which may include the contents of your files.
The results of tools the agent runs — file contents, search results, command output.
The API key for the model provider you configured.

Where it goes: to our service, which calls the model provider you chose and returns their answer to you. A few passes run the same way in the background, on the model you chose: naming a conversation, and — with Semantix Memory on — filing memories and writing the note the next session starts from.

With Local only on, we serve the request and keep nothing:

We do not store your prompts, your code, or your tool results.
We do not store your API key. It is used to make your call and it is not retained.
We do not train any model on any of it.
We do not sell it or share it with anyone other than the provider you chose.
Transmission is encrypted.

What the provider does is between you and them. OpenAI, Anthropic, Groq, OpenRouter, whoever you configured — their handling of your data is governed by their privacy policy, not this one, and it is worth reading.

3. What we keep when you allow it#

Each of these is a switch in the Studio's Settings → Privacy. Each one is off until you turn it on, and none of them is needed for the editor, the index, or the graph.

Local only, turned off. For agent runs in a project, we keep:

the conversation — its title and any instructions you gave it,
each run: its status, the model, the system prompt and agent settings, when it started and ended, and any error,
the messages, and each tool call with its input and its result,
a record of each step the agents took, used to find and fix problems,
token counts and cost for each model call.

We keep this so your run history and usage are there when you come back to them.

Back up conversations. A copy of each conversation, so it follows you to your other computers. Turning Local only back on offers to delete every backup we hold.

Semantix Memory. For your agents to remember across sessions, we keep:

the memories your agents write, and their links to each other,
a capture of each finished turn, which memories are filed from,
the note each session leaves for the next one,
the standing instructions an agent holds about its own conduct.

To make memories searchable, the text of each memory is turned into numbers (an "embedding"). We do that through OpenRouter, which sends the text to Nebius or DeepInfra to compute it. They receive each memory's text, and the text a recall searches with, for that purpose only.

Tasks. A task's schedule and settings, each run's state, who or what started it, when it ran, the last part of its output, and the computer it ran on (an identifier we generate, and when it was last seen). This is how your tasks show the same history wherever you open them, and how a plan's task limits are counted.

4. This website and your Semantix account#

If you create an account, we collect:

Your name and username — so the product can address you.
Your email address — to identify your account, to let you sign in, and to reach you about your account or important changes to the service.
Your password — stored as a cryptographic hash. We cannot read it, and neither can anyone who obtains the database.
If you sign in with Google or GitHub instead, the account ID, email address, display name, and profile picture link they send us.

To keep you signed in and secure, we keep each sign-in session with the IP address and the browser or app it came from, and the access tokens the Studio uses, stored as hashes.

Your account also holds the things you set up in it: your projects (their names and their folder paths on your computer), the agents you create, your privacy choices and when you made them, notes your agents leave each other, and the MCP servers you register, whose secrets are stored encrypted.

We keep a session cookie so you stay signed in between visits. It is strictly necessary for the site to work; it is not an advertising or tracking cookie, and we do not sell or share it.

Our servers keep standard technical logs — IP address, timestamp, requested page — as any web server does, for security and debugging. These are kept for a limited period and are not used to build a profile of you.

5. Payments#

Paid plans are sold by Paddle, our online reseller and the Merchant of Record for all our orders. When you buy a plan, you give your payment details, billing address, and email to Paddle, not to us, and Paddle's privacy notice covers what it does with them. We never see or store your card number.

Paddle tells us what we need to give you the plan you paid for: your Paddle customer ID, the subscription's status, its price, and when the current billing period ends. We keep that with your account.

6. What we do not do#

We do not sell your personal data. Not to anyone, at any price.
We do not share it with advertisers or data brokers.
We do not use third-party advertising or cross-site tracking on this site.
We do not train any model on your data.
We do not look at what we keep for you except to run the features you use, to fix a problem you report or one that breaks the service, to keep the service secure, or when the law requires it.

7. Who else is involved#

We keep this list short on purpose, and we will keep this page honest as it grows.

Hosting. Our services run on servers we rent, and this website is hosted by Vercel. Hosting providers process data on our behalf, under contract, and do not use it for their own purposes.
Model providers. Whoever you chose. You have a direct relationship with them, and your request reaches them because you asked us to send it.
Paddle — payments, tax, receipts and refunds, as the Merchant of Record (section 5).
OpenRouter, Nebius and DeepInfra — embeddings for Semantix Memory, only when it is on (section 3).
Google and GitHub — only if you sign in with them.
Cloudflare — our domain's DNS, and the delivery of email you send to our addresses.

If we ever add a service that processes your personal data, we will name it here before it goes live.

8. Your rights#

Wherever you live, you may ask us to:

See what personal data we hold about you.
Correct anything that is wrong.
Delete your account and the data attached to it.
Export your data in a portable form.
Object to a particular use of it.

You can also change your mind at any time: turn a switch off in Settings → Privacy, and from then on we keep nothing new under it.

Write to privacy@semantixstudio.com and we will act on it within 30 days. We will not make you justify the request, and we will not make deletion difficult.

If you are in the European Economic Area or the United Kingdom, the GDPR gives you these rights as a matter of law, and you also have the right to complain to your national data protection authority. Our legal basis for your account, a plan, and the features you use is the performance of our contract with you. For what the Privacy switches keep, it is your consent, which you can withdraw by turning them off. For security logs and sign-in sessions, it is our legitimate interest in keeping the service and your account safe. Billing records are kept because the law requires it.

9. How long we keep things#

Agent requests with Local only on — not retained. Prompts, code context, tool results and keys are used to serve the request and are not written to storage.
Agent runs, backups, and memories — until you delete them or your account.
Tasks and their run history — for as long as your account exists.
Account data — for as long as your account exists, and deleted when you delete it.
Sign-in sessions — until they expire or you sign out.
Billing records — for as long as tax and accounting law requires.
Server logs — a limited retention period, then discarded.
Backups — deleted data may persist in encrypted backups for a short window before those backups rotate out.

10. International transfers#

We are in Israel, which the European Commission recognises as giving personal data adequate protection. Some of the providers in section 7 process data in other countries, including the United States. Where the GDPR or the UK GDPR applies, those transfers rely on an adequacy decision or on the standard contractual clauses the provider has agreed to.

11. Children#

Semantix is not directed at children under 16, and we do not knowingly collect their personal data. If you believe a child has given us data, write to us and we will remove it.

12. Security#

We take reasonable technical measures to protect your data: passwords and access tokens are hashed, secrets you give us are encrypted, connections are encrypted in transit, and access to production systems is limited.

No system is perfectly secure, and we will not pretend otherwise. If we ever suffer a breach affecting your personal data, we will tell you, and we will tell you what actually happened.

13. Changes to this policy#

The Studio is early software and our infrastructure is growing. When what we collect changes, this page changes with it — with a new version number and date at the top, before the change takes effect, not after.

We will not quietly widen what we collect and leave the old promise on the page.

14. Contact#

privacy@semantixstudio.com

Semantix, Israel.