Orca

  • ID: orca
  • Store: ~/Library/Application Support/orca/agent-session-journal.db, ~/.config/orca/agent-session-journal.db, %APPDATA%/orca/agent-session-journal.db
  • Read override: DEJA_ORCA_DB or DEJA_ORCA_ROOT
  • Format: SQLite (agent_session_records, journal_sessions, journal_rows)
  • Needs: sqlite3

Orca (stablyai/orca) is a desktop workspace manager for coding agents that

records session activity in agent-session-journal.db (and dev builds under

orca-dev). Each recorded session has a metadata row in agent_session_records

(session_id, record_json with cwd, workspaceId, title, createdAt,

updatedAt) and structured journal items in journal_rows (session_id,

epoch, seq, ts, row_json).

Because Orca streams revisions per itemId, the reader keeps the highest

revision for each item in sequence order and parses message, command,

and system item bodies along with embedded text, tool-call, and

tool-result blocks.

  • Wiring: deja install orca writes the mcpServers.deja block into

<OrcaRoot>/mcp.json, installs the skill at

<OrcaRoot>/skills/deja-history/SKILL.md, and writes the /deja command to

<OrcaRoot>/commands/deja.md.

  • Auto-recall: deja install orca-auto adds lifecycle hooks (SessionStart,

UserPromptSubmit, PreToolUse, PostToolUse, PreCompact, SessionEnd) and

statusLine into <OrcaRoot>/settings.json.

  • Resume: orca --resume <id> in the session's recorded directory.

Last verified: 2026-11-12

deja reads this format and forty-six others, and turns what it finds into memory your agents can search. See the harness matrix for what is wired where, or install it and search your own history.

Found this useful? Star deja-vu on GitHub.