Keep memory between runs
Give an agent a notebook that survives between runs — remember a fact once, recall it by meaning later, and forget it when you want it gone.
Durable memory is opt-in. Until it is turned on, Kortecx keeps nothing between runs.
An assistant that forgets everything between sessions never gets better at your work. Memory gives it a notebook: you can write in it, read it back, and tear a page out.
Turn it on
Memory has to be enabled on the runtime. It is off until it is configured, so nothing is kept by accident.
An embedding model has to be available. Recall finds memories by meaning rather than by keyword, which needs a model to turn your text into a vector. See Local inference for the model side.
Whose memory is it
Every memory is scoped to you — the identity the runtime derived from how you authenticated. A client cannot ask for another identity's memories, and it cannot claim to be someone else: the caller proposes, the runtime decides.
Write, read, forget
In the console you can write a memory in your own words, search the notebook by meaning, and read back everything that is in it, newest first.
Each entry shows what was remembered and when, and you can forget one from the same view — the console asks first, because forgetting erases it.
The notebook you see here is the one your agents read from, so what you add or remove takes effect on the next run.
Ageing memories out
A notebook nobody prunes becomes noise. Decay removes entries that are both old and rarely recalled — a fact you keep coming back to is protected by how often you use it, however old it is.
Decay previews by default. It shows you the candidates and evicts nothing until you say so.
Eviction is reversible. What was evicted is still there to look at, so you can see what went and bring it back.
How agents use it
Inside an agent run, memory shows up as tools the agent can call in its loop: one to remember, one to recall, one to list what it remembered. Recall is read-only and reaches nothing beyond your own notebook — no network, no files — so it does not need a person to approve it.
They also fail soft. If no model is available or the notebook is empty, the agent gets an honest empty answer it can reason about, rather than an error that ends the run.
Recall is recorded. The runtime keeps exactly which memories were read, in order, so you can go back afterwards and see precisely what the agent had in front of it at each step. Similarity scores are shown for your benefit only — they never decide anything.
Memory or a dataset?
Both find things by meaning.
- Memory is the agent's own notebook: small facts it writes for itself, scoped to you, ageing out over time.
- A dataset is a body of documents you ingest for the agent to read from — a manual, a policy archive, a folder of notes.
Use memory for what the agent learns. Use a dataset for what you already have.