Skip to content

Docs

Team and Enterprise

Operational memory

Operational memory — your workspace's "org brain" — distills everything that has happened into a searchable corpus: closed terminal & database session summaries, resolved incidents (what broke and how it was fixed), and operator notes. Instead of asking a teammate "have we seen this before?", you ask the memory.

How it fills up

A background indexer folds new material in automatically — there's nothing to wire up. It starts from the moment the feature is enabled (it doesn't backfill old history), then keeps current:

  • Sessions — once a terminal or database recording closes and is summarized, its summary becomes a memory entry.
  • Incidents — when an incident is resolved, its title, detail and outcome are folded in.
  • Notes — anything you add by hand (see below).

Searching & Recall

Open Knowledge from the sidebar and pick the Operational memory tab. Two ways to use it:

  • Search — type an error message, symptom, host or topic and get the most relevant records, newest first. Click one to jump to its source (a session playback or the incident).
  • Recall — ask a plain-language question ("how did we fix the payment worker stall?") and get a short synthesized answer with the records it drew from cited beneath it. Recall needs AI enabled for your workspace.

Adding a note

Capture a fix or a runbook snippet with Add note — a title and body. Notes are searchable alongside the derived corpus, so a "how we recovered X" you write today surfaces the next time someone hits X.

In the assistant

Runa can search operational memory too — ask it "have we hit this error before?" and it calls the operational_memory_search tool, grounding its answer in your real past work.

Permissions

memory.read memory.manage

memory.read to search, view and Recall; memory.manage to add notes. The page is hidden until an administrator enables operational memory for your workspace; Recall and the assistant tool additionally need AI enabled.