# Stay on the same page

Introducing brainplug

brainplug plugs into your agent’s brain and observes its every move. So when you and your agent drift, you can go back to the page where you were.

_A conversation is a path._

1. Work with your agent as you do.
2. It keeps the points you both agreed on.
3. Say back, and it returns to the last one.

**Why we are building it**

brainplug plugs into your agent’s brain so you can see it act around what you ‘ask’.

Agents now work for hours on what we ask. We can’t understand what agents think until we can see it. In our experiments, brainplug found the right chain of thought after 29 compactions in a single session. That session ran to 2.37 billion tokens, and not a word of it was lost.

**The result so far**

- Fresh agent: Tokens 52.1k; Time 5.7 s; Message found? -

_“I think we have drifted apart. Get us to the same page.”_

- Same agent, Using brainplug: Tokens 52.7k (+609); Time 11.3 s; Message found? yes
- Same agent, Without brainplug: Tokens 67.6k (+15.5k); Time 28.4 s; Message found? yes

+ compared to the fresh agent

- 47×: a session file can be this many times a context window

## Request early access to

We're rolling it out to more people.

[Request access](mailto:hello@household.dev?subject=Request%20early%20access%20to%20brainplug)

[Other research](/research)
[Other tools](/#make)

## For agents

- [llms.txt](https://www.household.dev/llms.txt)
- [tools/index.json](https://www.household.dev/tools/index.json)
