Skip to main content
Last updated: 2026-09-06
orca memory inspects agent memories and the memory bank that spans every profile in the tenant. Memories are written by agents as they run (or extracted in the dashboard), so the CLI is read, search, and delete only; there is no create or update. Every subcommand accepts the global flags. delete prompts for confirmation in a terminal; in a script --yes is required or the command refuses (exit code 2).

​
orca memory list

Lists an agent’s memories. These are the command’s own paging flags, not the shared page flags: there is no --all, and neither flag has a CLI default. When omitted they are not sent and the server pages with its own values. When the server holds more memories than the page returned, the terminal view prints a Showing X of Y hint on stderr.
Plain (piped) output is one tab-separated row per memory: id, category, source, confidence, createdAt, summary. The --json output is the array of memory records only, without the page totals.
Searches an agent’s memories by relevance. Both flags are forwarded only when set; the defaults shown are applied by the server.
Plain (piped) output is one tab-separated row per hit: id, score, category, summary. The --json output is the array of scored results; each has memory (the full record) and relevance (score, recency, usage, topic).

​
orca memory show

Shows one memory in full: category, source, confidence, staleness, access count, active flag, version, timestamps, the summary, the processed content, and the raw input when it differs from the processed content.
Plain (piped) output is one field, value row per field: id, profile, category, source, confidence, staleness, accessCount, isActive, version, createdAt, lastAccessedAt, summary, processedContent. With --json the full memory record is printed.

​
orca memory delete

Deletes one memory.

​
orca memory bank

Prints a snapshot of the whole memory bank across every profile. The terminal and plain views flatten the snapshot into one list sorted by profile name, newest memory first within each profile, with a header that counts memories and profiles. With --json the snapshot is printed as the server returns it: total, profilesWithMemory, and memories, a map from profile name to that profile’s memory records.
Plain (piped) output is one tab-separated row per memory: profile, id, category, summary, createdAt.

​
orca memory bank stats

Shows bank-wide totals and a per-profile count, sorted by count descending.
Plain (piped) output is one tab-separated profile, count row per profile. The --json output is the stats record: totalMemories, profilesWithMemory, and perProfile, a map from profile name to count.
Verified against orca-cli cli-v0.5.0.