Add Supermemory to your AI agent

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Use DELETE_CONNECTOR for v5 namespace-scoped connector deletion instead. Delete a specific connection by ID. By default this also deletes documents imported by the connection; set deleteDocuments to 'false' to retain them. Returns deleted connection IDs and providers.

Disconnect a connector in a namespace and stop its webhooks. Imported documents are also deleted by default. Set deleteDocuments to false to preserve those documents.

Deprecated by the provider: the legacy v3 API shuts down on December 31, 2026. Use v5 DELETE /ns/{namespace} to delete a namespace instead. Delete a container tag and all its documents and memories. Only organization owners and admins can perform this destructive action.

Permanently delete one document by ID. This does not remove the document's container-tag settings record.

Retire a namespace and permanently delete its documents and memories. Optionally set moveTo to preserve its content by moving everything into another namespace first. A queued response only acknowledges an asynchronous move, not completed deletion; a deleted response confirms permanent removal.

Permanently delete 1 to 100 documents and their derived knowledge from a namespace by document ID or caller-defined ID. Each identifier is handled independently; successful deletions are preserved if others fail. Inspect count and every errors entry: HTTP 200 does not mean all documents were deleted.

Remove 1 to 50 profile bucket categories created specifically for a namespace and return its updated effective bucket definitions. Unknown names are safely ignored. Organization-wide buckets are protected and cannot be removed with this action. Requires namespace write access.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Use DELETE_CONNECTOR for v5 namespace-scoped connector deletion by connector ID instead. Delete connections for a specific provider and container tags. Returns the deleted connection IDs and providers.

Deprecated by Supermemory: legacy v3/v4 APIs shut down December 31, 2026. Use v5 DELETE /ns/{namespace}/memories/semantic for semantic forgetting, or DELETE /ns/{namespace}/memories for exact IDs. Soft-delete memories in one container matching a natural-language instruction or exact IDs. Set dry_run=true to preview first; omission defaults to real forgetting. Semantic matches may change between preview and apply; use exact IDs for a reviewable selection.

Deprecated by the provider; v3/v4 APIs shut down on December 31, 2026. Use v5 DELETE /ns/{namespace}/memories for new integrations. Soft-delete a memory by ID or exact content within a required container tag. The memory is marked as forgotten, not permanently deleted.

Remove 1 to 500 exact memories from normal recall in a namespace by stable memory ID, preserving their audit history. Each ID is handled independently: missing or ineligible memories are reported without rolling back successful changes. Inspect count, errors, and matches; HTTP 200 does not mean every requested memory was forgotten.

Preview or forget memories in a namespace using a natural-language query. Explicitly set dryRun to true to preview or false to apply forgetting. For a reviewable, drift-free workflow, preview here and pass the returned memory IDs to the exact-ID forget action instead of rerunning the query. Inspect count, matches, and errors for the provider-reported outcome.

Return the connected Supermemory organization's plan, credit balance, high-level usage, and reset date.

Deprecated legacy v3 API, scheduled to shut down December 31, 2026. Get the connected organization's automatic credit top-up configuration and payment-method availability without changing billing settings or purchasing credits. Requires an unscoped organization API key; scoped API keys cannot read billing.

Deprecated legacy v3 API; provider shutdown is December 31, 2026. Retrieve the connected organization's per-feature billing usage, limits, units, and period bounds when available. Requires an organization API key; scoped API keys cannot read billing. Console usage estimates are not authoritative invoices.

Deprecated legacy v3 API, scheduled to shut down December 31, 2026. Get daily billing meter events for the connected organization, including per-feature consumption and Memory versus SuperRAG token totals. Optional ISO 8601 start and end override the billing period. Requires an organization API key; scoped API keys cannot read billing. Console usage is an estimate; invoice line items are authoritative.

Deprecated by the provider: this legacy v3 endpoint shuts down December 31, 2026. Use the v5 GET /ns/{namespace}/connectors/{id} endpoint for new integrations. Retrieve a legacy connection by ID, including available provider metadata and the latest sync run status.

Read a connector in a namespace, including its setup, selection, and latest sync status. Optionally include the 10 most recent sync runs with failed items or a one-time hosted picker URL. Requesting picker requires admin write access because the link allows changing what syncs.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Use v5 GET /ns/{namespace} for namespace settings instead. Get a container tag's display name, entity context, memory filesystem paths, profile buckets, and creation and update timestamps.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Get the status and progress of a queued container tag merge by its merge job ID, including copied rows, renamed custom IDs, and the last error. This read does not start or retry a merge.

Get one document's content, metadata, source details, and processing status by document ID.

Return the ordered extracted chunks for one processed document.

Deprecated by the provider: this legacy endpoint shuts down December 31, 2026. Get a fresh presigned download URL for a document's uploaded file. The URL is valid for 24 hours. Returns an error if the document does not exist or has no file.

Deprecated legacy v3 endpoint scheduled to shut down December 31, 2026. Retrieve organization-wide error analytics, including request-type summaries, HTTP error breakdowns, hourly trends, and per-key metrics for a time window. Returns one page; increment page using the returned pagination metadata.

Read a namespace's supporting context and lifecycle timestamps. Supporting context helps Supermemory understand content and form better memories inside this namespace.

Retrieve a document's canonical content, metadata, and processing state by document ID or caller-defined ID within a namespace. Optionally include source chunks, derived memories, or both. Processing status done does not necessarily mean memory generation is complete; inspect dreamingStatus separately.

Retrieve one memory by ID within a namespace, including its text, metadata, version, and recall flags. Optionally include version history up to three steps away (nearest first), direct extends and derives links, and source documents. relatedLimit caps each related list separately.

Read the continuously maintained profile of a namespace's subject without a search query. Returns stable facts, evolving context, and custom buckets; optionally restrict contributing memories with metadata filters or select buckets.

Inspect the profile taxonomy active in a Supermemory namespace. Returns a name-to-description map combining organization-wide profile buckets with namespace-owned additions.

Read the connected organization's shared background that guides memory formation across all namespaces, together with its active namespace count. This action takes no parameters and does not change organization settings.

Retrieve the static, dynamic, or bucketed profile learned for one container tag, optionally with matching search results.

Deprecated by the provider: this legacy v4 endpoint shuts down on December 31, 2026. Use v5 GET /ns/{namespace}/profile/buckets instead. Return the effective profile bucket definitions for a container tag, merging organization-level buckets with container-tag-level additions.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Use v5 LIST_CONNECTORS to find namespace connectors and GET_CONNECTOR to retrieve one by ID. Retrieve a legacy connection by provider and container tags, including its latest synchronization status when available. This POST only reads connection details.

Deprecated legacy v3 API, scheduled to shut down December 31, 2026. Use the v5 organization and namespace settings APIs for new integrations. Retrieve the connected organization's ingestion filters, connector OAuth configuration, chunk size, and profile buckets. Connector client secrets are omitted from the response.

Deprecated legacy v3 API, scheduled to shut down December 31, 2026. Get organization-wide request usage analytics, including counts by request type, hourly performance, per-API-key metrics, and total memories. Filter by a relative period or ISO 8601 from_ and to bounds. Returns one page; use pagination to request further pages. Recorded tokensUsed is not the billable ingestion-token meter.

Deprecated by the provider: the v3/v4 API shuts down December 31, 2026. For new integrations, ingest a conversation transcript through v5 POST /ns/{namespace}/document instead. Ingest or update a conversation by its conversationId using structured messages, optional container tags, and metadata. Returns the backing document ID and processing status, not a guarantee that asynchronous processing has finished.

List one page of connectors across every namespace the API key can read, optionally filtered by provider. Use for admin views spanning users or projects. Pending connectors are not listed. Increment page up to pagination.totalPages to retrieve subsequent pages.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. For v5 document listing, use LIST_NAMESPACE_RESOURCES with type documents; v5 connectors are namespace-scoped. List documents indexed for a connected provider, optionally filtered by container tags. Returns document IDs, processing statuses, titles, summaries, types, and timestamps.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. List one page of accessible resources for a connection; only GitHub is currently documented as supported. Use page and per_page to request additional pages, and parent_id to scope resources to a parent.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Use v5 LIST_CONNECTORS for namespace connectors or LIST_ALL_CONNECTORS across namespaces. List all legacy connections, optionally filtered by container tags, including the latest synchronization status when available.

List one page of connectors that sync into a namespace, optionally filtered by provider. Pending authorization connectors are not listed. Use page and limit with the returned pagination metadata to retrieve subsequent pages.

Deprecated by the provider: the legacy v3 API shuts down December 31, 2026. Use the v5 namespace listing API for new integrations. List all container tags, optionally restricted by a comma-separated filter, including summaries, Nova flags, document and memory counts, and active merge details.

List documents, optionally filtered by containers, metadata, or filepath, one page at a time.

List a namespace's inferred memories awaiting review. This review route is still served under v3; pass the namespace as containerTag. Returns up to 50 memories ordered by parentCount descending, then createdAt descending, excluding forgotten, expired, or reviewed memories. No pagination is documented; total can exceed the returned count. Unknown or empty namespaces return an empty list and zero total. No v5 replacement is documented.

Deprecated legacy v3 API, scheduled to shut down December 31, 2026. List the connected Supermemory organization's billing invoices without changing billing settings or purchasing credits. Requires an unscoped organization API key; scoped API keys cannot read billing.

List extracted memory entries for one or more container tags, one page at a time.

Browse one page of documents, extracted chunks, or learned memories in a namespace. The other collection arrays are empty. Use pagination.totalPages and page to retrieve further pages. Position sorting is for chunks only; include.content and include.memoryCount apply to documents, and include.forgotten applies to memories.

List namespaces available to the caller, newest first, including their purpose, document and memory counts, and lifecycle timestamps. Optionally filter names with q. Returns one page; use page and the returned pagination to fetch more.

Deprecated by the provider: the legacy v3 API shuts down December 31, 2026. Use the v5 namespace document listing API for new integrations. List the live processing queue, all unfinished documents, or unfinished and failed documents. For view=all, fetch subsequent pages using page and limit and inspect the returned pagination information.

Deprecated legacy v3 endpoint scheduled to shut down December 31, 2026. List one page of organization-wide request logs filtered by time. Use pagination metadata and increment page to continue. Outputs are compacted and may be truncated. The ordinary tracker skips HTTP 402, 404, and 500 responses, so logs are not a complete audit trail. Historical retention is not guaranteed. Request type, status code, and API key filters are not supported.

Deprecated by the provider: the legacy v3 API shuts down December 31, 2026. Queue a merge of exactly two container tags into a target tag. All source documents are updated to reference the target, and source tags are deleted after a successful merge. Requires an organization admin or owner. The response confirms queuing, not completion; use GET_CONTAINER_TAG_MERGE with the returned mergeId to check progress.

Replace an existing document with a new file within an isolated namespace, keeping its stable document ID. Content and caller metadata are overwritten and the document is reprocessed asynchronously. The returned status reports processing state, not guaranteed completion. Requires namespace write access.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Destructively reset the ENTIRE connected organization's content: delete documents, memories, spaces except the default project, connections, and organization settings, and clear the default space context. Preserves the organization itself, members, and billing. This is not scoped to an individual document, container tag, or namespace. Only call with explicit authorization to delete the entire organization's data.

Provider deprecation notice: legacy v3/v4 APIs are scheduled to shut down December 31, 2026. These review routes still serve namespaces under v3; no v5 replacement is documented. Approve, decline, or undo a review of one inferred memory in a namespace. Approve confirms the fact; decline forgets it and removes it from search; undo restores the unreviewed inferred state, including un-forgetting a declined memory. Requires a key with write access to the namespace. A non-reviewable memory or an undo without a prior review returns 409.

Deprecated by the provider: this legacy v3 endpoint shuts down on December 31, 2026. Revoke a scoped API key using the ID returned at creation. Subsequent requests using the revoked key receive 401. Memories and namespaces are not deleted. Authenticate with an organization API key, not a scoped key.

Semantically search Supermemory memories, document chunks, or both, with optional container and metadata filters.

Deprecated by the provider: legacy v3/v4 APIs shut down December 31, 2026. Use v5 namespace search (POST /ns/{namespace}/search) for new integrations. Search documents and their content chunks with optional container, document, metadata, and filepath filters. Returns one bounded result set with relevance scores and optional full content or summaries.

Recall relevant learned context and source passages from an isolated namespace. Hybrid search combines memories and document chunks by default. Supports typed metadata filters (up to five levels), query rewriting, reranking, and optional source documents and related memories.

Send a bug report, feature request, documentation problem, or other feedback to the Supermemory team. This contacts real engineers and consumes the organization's quota of three submissions per UTC day, subject to a shared authenticated pool of 200 per day. Do not automatically retry uncertain deliveries, including HTTP 502: delivery may already have occurred.

Deprecated by the provider: the legacy v3 API shuts down December 31, 2026. Suggest profile bucket definitions using the organization's saved context prompt. Returns 3 to 6 suggestions tailored to that context without saving bucket definitions. Fails when no organization context is available.

Deprecated provider v3 API; shuts down December 31, 2026. For v5, use SYNC_CONNECTOR with a namespace and connector ID instead. Initiate a manual sync of connections for a provider, optionally filtered by container tags. The acknowledgement means importing was accepted, not that sync has finished.

Start a sync now for a connector in a namespace. Returns the connector ID and queued status, not a completed sync result. Returns a conflict if a sync for this connector is already running. Use Get Connector to inspect sync progress and results.

Change what a connector syncs in a namespace or its maximum document count. A supplied selection replaces the old selection and starts a sync. Only send selection for providers that support choosing resources. Returns the updated connector and its latest sync status.

Deprecated by the provider: the legacy v3 API shuts down December 31, 2026. Use the v5 namespace settings API (PATCH /ns/{namespace}) for new integrations. Update an existing container tag's display name, document processing context, memory filesystem paths, or tag-level profile buckets. Omitted fields remain unchanged; explicit nulls are sent for fields that support them.

Update the content, metadata, container, or processing configuration of one existing document. When metadata is supplied, it completely replaces the document's metadata, so omitted metadata keys are removed.

Deprecated by the provider; v3/v4 APIs shut down on December 31, 2026. Use v5 namespace document updates for document-backed memory changes in new integrations. Update a memory by ID or exact content within a container, creating a new version. The original memory is preserved with isLatest=false. Omitted metadata and expiry inherit from the previous version; explicit null forgetAfter clears the expiry.

Update the background context used during ingestion and memory formation in an existing namespace. Omit supportingContext to leave it unchanged, or explicitly set it to null to clear it. Requires write access.

Partially update an existing document in an isolated namespace without changing its stable ID. Supplied content replaces canonical content and is reprocessed; omitted fields remain unchanged. Returns the document ID and processing state, not a guarantee of completed processing.

Partially update a file-backed document in an isolated namespace. Supply a file to replace canonical content, or update metadata and processing context without a file. Omitted metadata and processing context remain unchanged. Returns the document ID and current processing state, not a guarantee of completed processing.

Update organization-wide background that guides content interpretation and memory formation across every namespace. Provide organizationalContext as a nonempty string to set it, or null to remove the existing context. Requires admin access.

Deprecated by the provider: the legacy v3 API shuts down December 31, 2026. Use v5 organization and namespace settings APIs for new integrations. Update the connected organization's settings, including filtering, custom connector OAuth credentials, chunk size, and profile buckets. Omitted fields are unchanged; explicit nulls are sent to the provider. This changes live organization configuration. Connector client secrets are accepted as input but omitted from the response.

Deprecated by the provider: this legacy endpoint shuts down December 31, 2026. Use the v5 namespace file upload endpoint POST /ns/{namespace}/document/file for new integrations. Upload a file as a document for processing, with optional container tags, metadata, and extraction settings. The returned status reports processing state and does not guarantee processing has finished.

Upload a file into an isolated Supermemory namespace to form searchable knowledge and learned memory. Ingestion returns once safely queued; extraction and memory formation continue asynchronously. The returned status is the document's current processing state, not a guarantee that processing has completed. Use task_type=superrag for retrieval without generating memories.

Add or update namespace-owned profile bucket definitions that guide memory classification. Omitted buckets remain unchanged. Organization-wide buckets are protected and cannot be changed. Returns the effective definitions, including organization buckets. Requires namespace write access.