Take your memory anywhere.

Why exports matter

Memory in mnueron is a database, but the tools you use day to day expect files. When you want to brief a new chat session on the last six months of work, you don't want to retype a summary — you want to drop in your saved notes. When you want a human collaborator to see how a decision evolved, you want a single document, not access to a dashboard. Exports bridge those use cases:

• Reuse as agent context. Drop the markdown into Claude Desktop, Cursor, Cline, or any chat that accepts file attachments. The agent grounds its responses in your actual history instead of starting fresh.
• Share with humans. HTML opens in any browser; Print → Save as PDF gives you a polished read-only document you can email or attach to a ticket.
• Back up or migrate. JSON / NDJSON are round-trip safe — re-import into another mnueron instance to restore your namespace exactly. Useful before destructive operations or when moving between local-first and hosted modes.
• Analyze. Open the CSV in a spreadsheet, query the NDJSON with jq or DuckDB, or pipe it into your own data warehouse. Your memory is just rows you have first-class access to.

Group by conversation, namespace, or not at all

The group_by option controls how memories are clustered in the output. Picking the right grouping mode turns a raw memory dump into a document that actually reads well.

API reference

The dashboard dialog is a thin wrapper around the export endpoint. If you want to automate exports (nightly backups, CI workflows, custom UIs), hit the endpoint directly.

curl

curl -X POST https://your-mnueron.app/api/memories/exports \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -d '{
    "format": "md",
    "group_by": "thread",
    "namespace": "claude-cowork"
  }' > export.md

JavaScript (browser, logged in)

const resp = await fetch("/api/memories/exports", {
  method: "POST",
  credentials: "same-origin",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    format: "md",
    group_by: "thread",
    namespace: "mnueron-dev",
  }),
});
const data = await resp.json();
const blob = new Blob([data.payload], { type: "text/markdown" });
const a = document.createElement("a");
a.href = URL.createObjectURL(blob);
a.download = "mnueron-dev.md";
a.click();

Response shape

{
  "id":           "export-uuid",
  "status":       "completed" | "pending",
  "format":       "md",
  "row_count":    127,
  "filters":      { /* echo of what you asked for */ },
  "payload":      "# Memory export\n\n...",
  "download_url": null,
  "created_at":   1735503600000,
  "completed_at": 1735503601234
}

Round-trip: export, then re-import

JSON and NDJSON exports are intentionally round-trip safe. You can POST the file back to /api/memories/batch and every memory is restored with its original namespace, tags, source, source_ref, metadata, and timestamps. This makes exports practical as a backup format or a way to migrate between mnueron instances (e.g. from a hosted prod environment to a local-first SQLite store).

Markdown and HTML are not round-trip safe — they discard some structured fields in the name of human readability. Tag arrays become comma-separated chips, metadata flattens to the visible meta line, and timestamps round to minute precision. Use them as outputs for human / LLM consumption; use JSON / NDJSON for anything that needs to come back as memory.