Skip to main content

Data

The Data group covers what agents remember and what they produce. Memory is an admin surface (org_admin and above); File Space stays visible to every signed-in user, because agents' deliverables and download links land there. Sidebar items are role-filtered, so a non-admin sees this group with File Space as its only entry.

Memory

/memory ("Memory & Storage") manages agent memory namespaces, retention and the raw store. Requires org_admin or above.

What you see:

  • Stat cards for total namespaces, total entries and scope levels in use.
  • A System Tags card — a per-agent operational reference (channels, people, IDs, tools-for-purpose) that is loaded into every session automatically. Pick an agent, then edit the text directly; Insert template, Revert and Save buttons manage the draft. System agents (such as the Orchestrator) are only listed for platform-level admins.
  • A namespace list alongside an entries pane: click a namespace to browse its entries, delete individual entries, or delete the namespace (which removes all of its entries).

Key actions:

  • New Namespace creates a memory namespace with:
    • Scope — org, team, agent, project or user.
    • Vector Backend — pgvector or Qdrant.
    • Max Entries (0 = unlimited) and Retention Days (0 = forever).
    • Optional PII detection & redaction.

Notable behaviours:

  • Deleting a namespace is destructive and confirmed with a danger prompt — it takes every entry with it.
  • Retention and entry caps are enforced per namespace, so different scopes can carry different lifetimes.

File Space

/file-space is EnClaw's virtual file system — where uploads, agent outputs and generated documents live. It is visible to all signed-in users: when an agent hands you a download link in chat, it points here. Per-file access control is enforced server-side, so seeing the page never exposes files you don't have rights to.

What you see:

  • Stat cards for directories, files, active shares and mount points.
  • A Folders pane with the directory tree and a Show system toggle for system-generated folders.
  • A file table per directory: name, type, size, a PII column and per-file actions.

Key actions:

  • New Directory creates a folder with a scope of org, team, agent or user.
  • Drag files from your desktop into a directory to upload them.
  • Per file: Download, Share, Delete, plus copy-path and copy-ID shortcuts for referencing the file elsewhere.
  • Share creates links with Read or Read/Write permission and an expiry date (or never). Existing shares are listed with their status (active, expired, revoked) and can be revoked.
  • Storage Settings manages storage mount points. A directory can be backed by Local storage (not recommended for production), Azure Blob Storage (connection string + container) or Amazon S3 (access keys, bucket, region).

Notable behaviours:

  • Files flagged in the PII column show a "Detected" badge; clicking it reveals the detected PII types and where they were found, with a prompt to redact or restrict access.
  • Each directory has an access log (time, user, action, file, IP address) so you can trace who touched what.
  • Download links of the form /file-space/file/<id> open a small landing page that authenticates you and then serves the file — safe to paste into chat, email or documents.

For governance guidance (scopes, sharing policy, storage backends) see File Space.