GBrain: One Memory, Many Models, One Workspace Price
ProductivityPaidGBrain gives a team one shared memory folder that ChatGPT, Claude, Cursor and other AI clients read and write, priced per workspace rather than per person. Permissions, daily tool limits, access end dates and an activity log are all set by the team.
About GBrain
The billing unit is the first thing worth noticing. GBrain charges one price per workspace rather than per person, and inviting the rest of the team does not move it, so a three-person team and a thirty-person team cost the same. That is unusual among AI tools priced by seat, and it explains why the product is organised around shared context instead of personal assistants. What gets shared is a memory folder, described on the site as plain markdown files in a directory you can read, correct and copy. Fix a note once and every AI connected to the workspace picks up the correction on its next turn. When you leave, you copy the folder and the notes come with you, and the open-source parts are described as free to run yourself. A portability claim like that is unusual enough to be worth testing before you depend on it. The permission model has three layers. At the memory level, each AI is set to Off, Read or Full, where Read can search and recall and Full can also remember, link and forget. At the application level, permissions are granted per connected service rather than globally: Gmail has four levels including Draft and Manage, Calendar has two, and web search is on or off. Raising a level does not hand over a credential, because connected accounts are held by GBrain and an AI receives the result of a call rather than the key. Daily caps sit underneath, per tool per account, and each AI's access can be given an end date, after which its next call is refused. Repeating work is handled by skills, which are jobs the workspace knows how to run on demand or on a schedule. Give one a time and it runs, writes what it found into the memory and posts back to the thread. Team discussion happens in a single shared conversation with the assistant answering for everyone, plus named threads per piece of work with unread counts. The security documentation is the most candid part of the product, and it is where a careful buyer should start. Each brain runs on dedicated single-tenant compute with its own encrypted storage, isolated from other brains and not reachable from the public internet inbound, with the control plane in a separate environment. In the same document GBrain states that it holds no SOC 2 or ISO 27001 certification, that it offers no end-to-end encryption because a brain must read team content in cleartext to summarise it, and that SSO, SCIM and per-brain member permissions are not offered today. Processing happens in the United States, the default setup uses Anthropic for the model and OpenAI for embeddings, and the vendor states it does not train models on customer content. Read together, that is a coherent position rather than a list of gaps: strong isolation, and honesty about the certification and encryption limits. It is still not a finished compliance story. Anyone bound by a formal security review should raise the missing certifications and the absence of SSO before buying rather than after, and should scope a second look at the 30-day vulnerability fix commitment, which is a policy rather than a contractual guarantee.
Pricing & Plans
There is one plan. Explorer is $199 per month per workspace with $100 of usage credit included each month and no per-person charge. Billed monthly and cancellable from the workspace billing page.
Explorer
One price per workspace, whatever the team size
- $100 of usage credit included every month
- Models from Anthropic and OpenAI, switchable at any time
- Scheduled skills that run unattended
- The whole team in one multiplayer workspace
- Email, calendar and web connected once
- Onboarding and support direct from the vendor
Best For
Recommended use cases and scenarios where GBrain shines.
Pros
- Priced per workspace rather than per person, so inviting a team does not raise the bill
- Memory is plain markdown in a folder you can read, correct and copy out when you leave
- Credentials stay with GBrain, so an AI receives the result of a call instead of the account key
- Permissions are set per application and per AI, with daily tool caps and end dates on access
- Skills can run on a schedule, write what they found into the memory and report to the team thread
- The security documentation states its certification and encryption limits plainly instead of leaving them implied
Cons
- No SOC 2 or ISO 27001 certification, stated openly on the vendor's own security page
- No end-to-end encryption by design, because the assistant has to read team content in cleartext to summarise it
- SSO, SCIM and per-brain member permissions are not offered today
- Processing happens in the United States, with EU residency described as something to raise with them
- A single $199 monthly plan is a large commitment for a small team to experiment with
Reviews & Ratings
Loading reviews...
Similar Tools
More Productivity tools you might like
Zapier
No-code automation platform that connects 9,000+ apps with Zaps, plus AI agents and chatbots that work across your tools.
Make
Visual no-code automation and integration platform with a drag-and-drop scenario builder, routers, and a generous free tier.
