Skip to main content
GET
Get memory

Authorizations

Authorization
string
header
required

API Key or JWT User Token. If using an API Key, set the X-As-User header to act as a specific user. A JWT User Token is always scoped to a specific user.

Headers

X-Hyperspell-Response-Profile
enum<string>
default:api

Case-sensitive api or mcp. Channel pages use 32 KiB for both profiles; other reads use 1 MiB for api and 256 KiB for mcp.

Available options:
api,
mcp
X-As-User
string | null

Optionally set this header to act as a specific user when using an API Key, equivalent to first exchanging the API Key for a User Token

Path Parameters

source
enum<string>
required
Available options:
reddit,
notion,
slack,
google_calendar,
google_mail,
imap,
google_meet,
box,
dropbox,
github,
gitlab,
google_drive,
vault,
web_crawler,
trace,
microsoft_outlook,
microsoft_teams,
granola,
fathom,
fireflies,
figma,
linear,
hubspot,
salesforce,
coda,
confluence,
jira,
metabase,
gong,
clickup,
lightfield,
pylon,
fellow,
odoo,
external_mcp,
page
resource_id
string
required

Query Parameters

include_chunks
boolean
default:false

When true, include the document's extracted memories (chunks with summaries) in the chunks field, in document order.

cursor
string | null

Continuation from a Slack or Teams channel page.

Maximum string length: 128

Response

Successful Response

resource_id
string
required
source
enum<string>
required
Available options:
reddit,
notion,
slack,
google_calendar,
google_mail,
imap,
google_meet,
box,
dropbox,
github,
gitlab,
google_drive,
vault,
web_crawler,
trace,
microsoft_outlook,
microsoft_teams,
granola,
fathom,
fireflies,
figma,
linear,
hubspot,
salesforce,
coda,
confluence,
jira,
metabase,
gong,
clickup,
lightfield,
pylon,
fellow,
odoo,
external_mcp,
page
type
string
required

Hyperdoc document type discriminator (document, message, file, event, ...).

document
Document · object
required

The hyperdoc tree. Switch on type for the document frame and recurse through children for the body.

Example:
body_status
enum<string> | null

Content representation: full, matched excerpt, partial preview, or omitted. Absent on reads that do not yet report this status.

Available options:
full,
excerpt,
preview,
omitted
title
string | null

Human-readable document title.

status
enum<string> | null

Indexing status of the document.

Available options:
pending,
processing,
completed,
failed,
pending_review,
skipped,
filtered,
cancelled
collection
string | null

The document's collection, if any.

metadata
Metadata · object

Filterable custom metadata attached to the document.

ingested_at
string<date-time> | null

When Hyperspell first indexed the document.

last_modified_at
string<date-time> | null

When the source document was last modified, if supplied by the source.

document_date
string<date-time> | null

The document's own date (e.g. email sent date, event date).

chunks
MemoryChunk · object[] | null

Extracted memories (chunks with summaries) for this document, in document order. Present only when explicitly requested via include_chunks; omitted otherwise.

next_cursor
string | null
notices
ReadContentNotice · object[] | null