What you get, and when you get it
Your agents' memory fills up as they work. Here is what lands on your side at each point, starting with the first hour.
- In the first hour
Your files, sorted and reviewed
The instruction files and memory folders your agents already use are imported. You settle what contradicts, approve the rest, and your agent starts its first session with memory. One hour from invite to that session is the bar we hold ourselves to.
- Every session
An agent that starts where the last one stopped
Its slice arrives before your first prompt. It keeps notes as it works, so a fresh session can pick up a task that was left half done. What it learns stays with it.
- Every week
A report on what your agents did with memory
Sessions, memories written, lessons proposed and what you did with them, contradictions, and every time an agent was refused something.
See a sample report - After a few weeks
A side-by-side on your own tasks
The same agent on the same tasks, once with memory and once without. Whether a task passed is decided by tests the agent never sees. Cost is metered from outside the agent. You get the write-up, and you sign it off.
See a sample side-by-side - Whenever you like
Everything you put in, back out
One command exports your team's memory. The lessons and decisions you approved are already written back into your own files, so you could stop tomorrow and lose nothing.
What you won't have to do
- Write memory by hand. It comes from the files you have and the work your agents do.
- Host anything. It's a service. You install one command-line tool.
- Change how your agents work. Same tools, same files, same prompts.
- Pay to start. Teams of up to 10 are free.
What we ask of you
- At least one agent that does real work every week.
- A few minutes of review a week. Drafts wait for you. Nothing is shared until you say so.
- Your honest verdict. If memory didn't help, we need to hear that more than anything.