Agents · Brain

Memory for your agents.

A space where your assistant saves what was decided, what you prefer and where the work stopped, and finds it again in the next conversation. You see, search and fix everything in a dashboard.

Sign in with a code sent to your email. No password, no key. The dashboard is in Portuguese for now.

Why this exists

You already explained this. Someone needs to
REMEMBER.

The decision, the preference and the point where the work stopped live in the conversation where they were said. In the next one, or in another assistant, someone has to say them again. In Brain, what counts lives in a space of yours: the assistant saves it when it happens and searches before acting, and you see, fix and delete it in a dashboard.

  • The same explanation, conversation after conversation
  • The decision nobody knows the source of anymore
  • Work that starts from scratch after a break

The whole path

Save, search, review.

And you see the whole space, in a dashboard.

Search

By meaning and by exact word. And it says when it found nothing.

The agent search only looks at what counts: active, not expired, not replaced. It searches by meaning and by exact word at the same time, so a build number or a name shows up too. When nothing is close, it says so instead of guessing.

search (example)
search   "which database do we use for orders?"

Store.Checkout.Database   decision · by agent
Checkout uses MySQL, because the client requires it.
source: client email

nothing close   says it found nothing, does not guess
two decisions   warns that they disagree, does not pick

Save

Decision, preference, observation, resume point or reference, each with who wrote it. When the agent writes a decision or a preference, it includes the source and the reason.

Change and review

What changed becomes a version, and nothing is gone for good. Two decisions that disagree, expired memories and rules without a source wait for you in a single list.

See it in the dashboard

A map links memories by area, and the search box looks by meaning, exact word and text. Lists load a little at a time.

How the assistant reaches it

One address per space.

Sign in, authorize, done: there is no key to paste.

The address

Add it to your client. The login opens on its own.

Each space has its own address, on the Connection and settings screen. You authorize that space, read only or read and write. There are six tools, and when it connects, the assistant receives the rules for when to search, save and stop.

terminal
$ claude mcp add --transport http brain-aurora https://brain.worker.myagents.click/mcp/<space id>

the client opens the login
you authorize this space
read · read and write

The limits

What It Does, and Where It Stops

The limits, stated before you need to ask.

Access

Email login

A code arrives by email; the first login creates the account.

Per account

1 free space

3, 9 or 99 on the plans. Each space is its own database.

Expiry

Observation 30 days, resume point 14

Decision, preference and reference do not expire by date.

What the model saves on its own

Depends on the client

In testing, decision, change, reference and resume point were saved on their own. Preference competes with the client’s own file-based memory.

Clients tested

Claude Code and claude.ai

Other MCP clients with OAuth should work, but were not measured.

Import

Only Brain JSON

It does not import from other products.

Pricing

The plan counts spaces: 1, 3, 9 or 99

Free

1 space, with no plan limit on memories.

No card and no expiry. Paid plans are per product, monthly or yearly (yearly gives 2 months free), and a limit only applies when you create the next one.

Create a free account

FAQ

The Questions, in the Order They Come Up

Still have a question?

Message us on WhatsApp and tell us what your assistant needs to remember. The person who answers wrote the code.

Agents · Brain

Memory for your agents.

A space of your own, with one address per space and a login. You see and fix everything in a dashboard.