Data processing agreement
Brainstack acts as your processor under GDPR. The agreement lists what we process, where and with whom.
Read the DPAUse the Claude, ChatGPT or Gemini plan your company already pays for.
Each one with the exact actions an agent can take in it.
All integrations30 minutes with us. Bring one job you never want to do again.
Brainstack gives every company its own private server, with its own database, its own vault for passwords and encrypted backups every 6 hours. No other company's agents run on it, and nothing leaves without the approvals you set.
They work on your server, on your schedule. Each run happens in its own sealed session, so one task can't read another's files.
Every client file, decision and source the brain keeps is stored on this server, in your own database.
A PostgreSQL database that holds only your company. Connections to it are encrypted, even inside the server.
The logins and tokens for Gmail, HubSpot or Stripe are kept encrypted in a vault on your server, apart from everything else.
What your team uploads and what agents produce, from PDFs to reports, sits in your server's own file store.
Agents reach the internet through one gateway that only opens to the tools you connected. Everything that goes out passes the same door.
Each customer has a separate machine with its own keys, its own database and its own chat bot. Nothing is shared between them.
A copy is encrypted on your server every 6 hours, signed and sent to separate storage. Your server can write copies but can't read or delete them.
Tap any part of the server
In a shared app, every customer sits in the same database. With Brainstack, your company gets a machine of its own.
One database and one set of keys for every customer. A bug or a leaked key can touch all of them.One server per company, each with its own keys. A problem in one stays in that one.
Nothing of yours sits in a table next to another company's.
The keys that open your data are made for your server and live on it.
Telegram, Slack and Discord connect through your company's own bot, not one shared by every customer.
Your company gets a private server that no other customer touches. Here is exactly what stays there and what goes out, and why.
Client files, decisions, sources and knowledge spaces.
In the vault, encrypted, apart from the rest.
Uploads, generated documents, chat history and the activity log.
What an agent or a person sends to Claude, ChatGPT or Gemini is processed by that provider, under the terms of your plan with them.
An email to Gmail, an invoice to Stripe, a message to Slack. Only after the approvals you set.
Encrypted before they leave, so the storage only ever holds sealed copies.
You connect a tool once and its login goes into the vault. Agents use the connection when a job needs it. Tap a connection to follow one use.
Each copy is encrypted before it leaves your server and signed, so a restore can prove it's yours and that nothing is missing. Tap any copy.
Once a month a fresh copy is restored as a test and every table is counted against what was saved.
The key that opens a copy is never on your server, so the copy is useless to anyone who finds it.
Your server can add copies but has no key that deletes them. Storage keeps each copy for its full period.
Each copy carries a signature and a count of every table, so a restore proves it got everything.
Set access per person and per space. The same rule applies to their agents and to the AI they plug in.
| Space | Access | What their AI and agents can use |
|---|---|---|
| Clients | OpenRead onlyHidden | |
| Finance | OpenRead onlyHidden | |
| Playbooks | OpenRead onlyHidden | |
| Team and payroll | OpenRead onlyHidden |
Who did what, when, and who said yes. People and agents land in the same log, and you can export it whenever you need it.
You decide, per agent, what needs a person: sending, publishing, paying, deleting, or any amount above a limit.
Invoice #1043 for Northwind is ready: $4,200. Send it?
Above your $3,000 limit, so it waits for you. Sent. Logged with your name. Held. Nothing was sent.A data processing agreement on every plan, a full export whenever you ask, and a clean exit if you leave.
Brainstack acts as your processor under GDPR. The agreement lists what we process, where and with whom.
Read the DPADownload a readable export of your workspace through a private link that expires after a short time.
When you cancel, you get your export first. After the grace period, the server, its firewall and its address are deleted.
| Company | What for | Where |
|---|---|---|
| Hosting provider | Your private server, used only by your company | Close to your company, in the US or Europe |
| Amazon Web Services | Encrypted backups (sealed copies only) | Separate storage, sealed before it leaves |
| Stripe | Billing | Per Stripe's terms |
| Your AI provider | What agents and people send it, under your plan | Per your plan's terms |
On a private server that only your company uses. Your database, your vault and your files live there, apart from every other customer. We place it close to your company, in the US or in Europe. Encrypted backups go to separate storage. Your AI provider processes what an agent or a person sends it, under the terms of your plan with them.
Not yet. What we do have today: one server per customer, a vault for passwords and tokens, encrypted and signed backups with a monthly restore check, permissions per person, approvals before anything leaves, an activity log and a data processing agreement under GDPR.
No. Brainstack does not train any model. Agents use the Claude, ChatGPT or Gemini plan your company already has, and what that provider does with what it receives is set by your plan with them.
No customer shares your server, your database, your vault or your chat bot. Each company runs on its own machine with its own keys, and backups of each server are encrypted and stored apart.
A copy every 6 hours. Each copy is encrypted before it leaves the server and signed so a restore can prove it is yours and complete. Frequent copies are kept for two days, daily copies for two weeks and monthly copies for a year. Once a month a copy is restored as a test and every table is counted.
Only what you allow. Sending, publishing, paying and deleting can each require your approval, and amounts above a limit you set always wait for a yes. Email replies wait for approval by default.
Yes. You can download a readable export of your workspace through a private link that expires. When you cancel, you get the export first, and your server is deleted after the grace period.
Yes. Every plan includes a data processing agreement under GDPR. Our privacy policy lists every sub-processor: the hosting providers that run your server, AWS for encrypted backups, Stripe for billing, and the AI provider you choose.
Every plan comes with its own private server. From $99 a month, plus the AI plan your company already pays for.