Mnueron groups related memories into threads via metadata.parent_ref. The Chrome extension captures one memory per chat turn and stamps every turn with the same parent_ref, so a long conversation can be reassembled in order.
GET /api/threads
List threads — one row per parent_ref value, with a summary count and time range.
curl "https://www.mnueron.com/api/threads?namespace=claude-web&limit=50" \
-H "Authorization: Bearer $MNUERON_API_TOKEN"
Query
| Param | Default | Notes |
|---|---|---|
namespace | — | Filter threads to one project |
limit | 100 | 1..500 |
offset | 0 |
Response
[
{
"parent_ref": "chat-abc123",
"namespace": "claude-web",
"count": 14,
"first_at": 1737070000000,
"last_at": 1737070500000,
"title": "Refactor the chunker",
"has_chunks": true
}
]
Memories without a parent_ref show up as solo threads (count: 1).
GET /api/threads/:ref
Returns every memory in one thread, ordered by metadata.chunk_index (when present) then created_at.
curl https://www.mnueron.com/api/threads/chat-abc123 \
-H "Authorization: Bearer $MNUERON_API_TOKEN"
Response
{
"parent_ref": "chat-abc123",
"chunks": [
{ "id": "01HQ…", "content": "...", "metadata": { "chunk_index": 0, "role": "user" } },
{ "id": "01HR…", "content": "...", "metadata": { "chunk_index": 1, "role": "assistant" } }
]
}
If parent_ref doesn't match anything in the threads table, the route falls back to looking up a single memory id — so the same URL works for both a thread reassembly and a one-shot fetch.