Skip to main content
Memory is how an agent carries something forward. A run that worked out why a service was slow can write that down, and a later run of the same agent finds it instead of re-deriving it. This page covers what a memory is, the spaces that hold them, the review every write passes through before anyone can read it, and what you see on the Memory screen.
Memory is per agent, not per workspace. An agent reads only the spaces it is bound to, so a memory written by one agent is invisible to another until you bind it.

Spaces hold memories

A space is a named collection — Production incidents, say — and you bind agents to it. Binding is the whole access model: an agent with no space bound to it gets no results from a memory search, and an agent bound to two spaces searches both. The Memory screen gives you Manage spaces to create and bind them, Manage memories to read and retire individual entries, and Activity for what the review pass has been doing.

Every write is reviewed before anyone can read it

This is the part worth understanding, because it is not what “the agent saves a note” suggests. The quarantine exists for a specific failure: an agent that learns one lesson often saves it three slightly different ways in the same run. Holding every write as a draft means nothing is published until something has compared the three.

What happens after a run

Three stages, and each one is an ordinary run you can open and read.
1

The agent is asked what to remember

The same agent resumes a copy of its own thread and is asked what should be remembered. It answers with the full context of the work it just did, rather than a summary of it. Because the thread is copied, the conversation you were having is never touched.
2

The drafts are reviewed

A reviewer is handed the drafts, the run’s transcript, and any existing memories the drafts may duplicate — so the decision is made with the alternatives in view.
3

The decision is applied

Each draft is promoted, rewritten, merged into an existing memory, or rejected. Activity reports the outcome plainly: 2 created, 1 combined, 1 removed.

Writing the same thing twice writes it once

A write whose text already exists in any space the agent reads returns the existing memory rather than adding a second copy. Agents that re-learn the same fact do not inflate the space.

A coherence pass runs on the space itself

Every six hours a sweep asks a question no single run can: is this space still coherent? Two memories written weeks apart that say the same thing, one that contradicts another, wording that has aged badly — none of that is visible to a pass looking at one run’s output. A space holding fewer than two memories is skipped, since deduplication, merging, and contradiction all need something to compare against.

How a search finds a memory

A search runs two ways at once — keyword matching and meaning-based matching — and the results are fused into one ranking. Ranking is by each hit’s position in that fused list rather than by its raw score, because the two methods produce scores on scales an order of magnitude apart; ranking on the raw number put every meaning-matched memory above every keyword-matched one regardless of the evidence. Agents reach memory through two tools: one to search it, one to save to it.
Memories do not expire on their own. There is no time-to-live and no automatic retirement by age — retiring a memory is something you or the coherence pass does. Treat a space as something with an owner, not a cache that cleans itself.

Next steps

Knowledge base

Documents you upload, as opposed to what an agent learned on its own.

Knowledge graph

Explore what the fleet knows and how entities connect.

Runs & evidence

The digest and review stages are runs — open them like any other.

Skills

Know-how you author deliberately, rather than something learned.