Supermemory CLI
New command-line tool for managing memories, documents, profiles, tags, connectors, and API keys directly from the terminal.@supermemory/tools v1.4.1
Now exposes raw searchResults in MemoryPromptData, giving full control over how retrieved memories are formatted in prompts.PPTX & Audio Ingestion
PowerPoint files (.pptx) are now a supported content type. Audio files are automatically transcribed via Gemini 2.5 Flash, chunked, and indexed.Multi-containerTag Scoped API Keys
Scoped API keys can now be assigned to multiple container tags — one key, multiple spaces.Console: Documents Page
New dedicated documents browser in the console for viewing, filtering, and managing all ingested content.Delete Scoped API Keys
NewDELETE endpoint to disable scoped API keys programmatically.supermemory-agent-framework Python Package
Official Python package for using Supermemory with Microsoft’s Agent Framework — memory tools and middleware out of the box.Interactive Memory Graph in MCP
The MCP server now includes an interactive graph visualization app for exploring memory connections from any MCP-compatible client.More Integrations
- ViaSocket — new integration guide for automation workflows.
- Plugin Auth Connect Page — OAuth-style connect page for Claude Code, OpenCode, and ClawdBot.
- OpenAI SDK Backfill — improved compatibility across TypeScript and Python SDKs.
Other
- Retry failed documents by re-submitting with the same
customId. - Delete connection without documents — disconnect a source without deleting synced content.
- Org-level overage toggle in billing settings.
- Copyable team invite link on the team management page.
extendsrelation type in memory graph for richer knowledge graph connections.- Bulk delete in the Nova app interface.
Self-Hostable Supermemory
Run the full Supermemory stack on your own infrastructure with Docker.Console v2
Complete redesign of the developer console with new navigation, improved billing, and a unified project view that merges consumer and developer organizations.No More 120 Memory Limit
The previous cap of 120 memories per container tag has been removed. Store unlimited memories.Supermemory Skill for Claude Code
Install withnpx skills add supermemoryai/skills — teaches Claude to proactively recommend and implement Supermemory when building AI apps. Includes TypeScript and Python examples.API Improvements
- Metadata filtering for profiles — target profile queries by metadata fields.
- List documents with multiple container tags — new
operatorparameter. - Deprecate
include: chunksin/v4/searchin favor of thehybridsearch mode. - Content deduplication in search results to reduce token usage.
Unified Organizations
Consumer and developer organizations merged into a single org type. All orgs can now access both Nova and the developer API.Credits-Based Usage Display
Billing now shows token usage in a credits-based format.Nova Spaces with Multi-Select
Spaces in Nova support multi-select, replacing “All Spaces” with scoped “Nova Spaces.”Scoped API Keys for Container Tags
Create API keys scoped to specific container tags for fine-grained access control per space.DELETE Endpoint for Container Tags
New endpoint to delete container tags and their associated document relationships.Container Tag-Level Context Prompts
Set custom context prompts per container tag to control how memories are extracted and summarized within each space.New Framework Integration Docs
Added guides for LangGraph, OpenAI Agents SDK, CrewAI, Agno, Mastra, LangChain, and Claude Code — covering all major AI agent frameworks.Entity Context & Authentication Docs
New docs on entity extraction, context enrichment, and comprehensive authentication examples (API key, OAuth, scoped keys).Plugin Authentication System
New auth system for external tool integrations, enabling secure plugin-to-API connections. Dedicated plugin page with auth flows for Claude Code, OpenCode, and ClawdBot.Enterprise Plan Support
Enterprise tier now available in the console.@supermemory/tools — Strict Mode
Strict mode support for OpenAI function calling, ensuring schema-validated tool calls.Hybrid PDF Pipeline
PDF extraction now uses Mistral OCR 3 with Gemini fallback for significantly improved accuracy on scanned documents and complex layouts.Halfvec Embeddings
Embedding storage optimized with half-precision vectors, reducing storage costs while maintaining search quality.Spaces Creation with Emoji
Create and customize spaces with emoji identifiers in Nova.Gmail Connector
New connector to sync Gmail threads into Supermemory. Threads are stored in R2 for reliable processing of large mailboxes.supermemory-pipecat Python Package
New SDK for Pipecat voice AI pipelines, including Gemini Live speech-to-speech support.@supermemory/tools — Prompt Templates
Customize how memory context is formatted in AI SDK integrations with the new promptTemplate option.Other
- Container tag filters in list and search endpoints.
- Pagination improvements across the console.
MCP 4.0
Major MCP server update with session configuration, project-aware tools on every init, and backward-compatible 3.0 support. Newcontext prompt for automatic user profile injection into AI conversations.S3 Connector
New connector to sync documents from Amazon S3 buckets, with console UI for bucket configuration.Memory Graph Revamp
Complete rewrite of@supermemory/memory-graph with improved visualization and performance.@supermemory/tools — AI SDK v5/v6
Now supports both Vercel AI SDK v5 and v6 with automatic version detection.Conversation Support in SDKs
supermemory (TypeScript) and supermemory-openai-sdk (Python) now support the conversations API for multi-turn chat with memory.MemoryBench
New open-source benchmark suite for evaluating memory systems, with documentation and CLI.Hybrid Search Mode
Newhybrid search mode in /v4/search combining semantic and keyword search for better recall on technical queries.Firecrawl Integration
Web crawling powered by Firecrawl for more reliable extraction of website content, with fallback support.Custom GitHub Credentials
Bring your own GitHub OAuth app credentials for the GitHub connector, enabling private repo access.API Key Expiration Emails
API keys now trigger email notifications before expiration.Connector Sync Logs
Connection syncs now produce detailed logs visible in the console.@supermemory/tools — Browser API Key Support
apiKey can now be passed via options instead of relying on process.env, enabling browser-based usage.Web Crawler Connector
New connector to crawl and index entire websites with configurable depth and URL patterns.@supermemory/memory-graph Package
New package for building interactive graph visualizations of memory connections, with a standalone playground.OpenAI Responses API Support
@supermemory/tools OpenAI integration now supports the Responses API.supermemory-openai-sdk — Python Middleware
New withSupermemory middleware for the Python OpenAI SDK, enabling transparent memory injection into OpenAI API calls.Browser Extension Webpage Capture
Chrome extension can now capture full webpage content with markdown conversion.Bulk Memory Optimization
Memory creation now uses bulk inserts for significantly faster batch ingestion.Enhanced Filtering
Newstring_contains filter type for partial string matching, ignoreCase option for case-insensitive operations, and improved negation support. Enhanced SQL injection protection.withSupermemory for OpenAI SDK
New withSupermemory wrapper for the OpenAI TypeScript SDK — transparent memory injection with automatic assistant response capture.Zapier & n8n Integration Pages
New integration guides for connecting Supermemory with Zapier and n8n automation workflows.@supermemory/tools — AI SDK withSupermemory
New withSupermemory language model wrapper for Vercel AI SDK that automatically injects memory context and captures assistant responses.Raycast Extension
New Raycast extension for quick memory access and addition from the macOS launcher.User Profiles API
New/v4/profile endpoint for retrieving AI-generated user profiles derived from memory interactions, with container tag scoping.Other
- DOCX support — Word documents can now be ingested.
- Project selection for connectors — assign Google Drive, Notion, and OneDrive connections to specific projects.
- Multiple models in consumer chat — model switcher with system prompt improvements.
- Organization settings — configure Supermemory behavior (chunking, extraction, memory limits) per org.
Forgotten Memories Search
Newinclude.forgottenMemories parameter in v4 search API to search through memories that have been explicitly forgotten or expired.Enhanced Delete API
DELETE /v3/documents/:id now supports both internal document ID and customId.API Terminology Update
Renamed “memories” to “documents” for developer clarity. New/v3/documents/* endpoints with full backward compatibility via automatic redirects from /v3/memories/*.Console Revamp
New console design with dark/light mode, org switcher, billing invoices, space selector with search, and memory list with multi-delete.Other
- New filters — revamped filtering UI in the console.
- Onboarding redesign — new step-based onboarding with code samples.
- Configurable chunking — set chunk size and algorithm per org.
Documentation v2.0
Complete rewrite with comprehensive API references, cookbook recipes, and production-ready examples for TypeScript, Python, and cURL.@supermemory/tools Package
New tools package for native Vercel AI SDK and OpenAI integration with memory tools and infinite chat. Plus openai-python-sdk for Python middleware.Batch Add & Bulk Delete
NewPOST /v3/documents/batch for batch ingestion and DELETE /v3/documents/bulk for bulk deletion.Memory Forgetfulness System
Full lifecycle management withforgetAfter dates and forgotten memory filtering.Video Uploads
Video files can now be ingested with automatic content extraction.MCP Connection Flow Redesign
Step-based UI for connecting MCP clients with v1 migration support. One-click install for Cursor.Claude.ai & t3.chat Extension Support
Browser extension now integrates directly with Claude.ai and t3.chat for automatic memory search during conversations.Waitlist Removed
Supermemory is now open to all users — no more waitlist.New Landing Page & Developer Page
Redesigned marketing pages with developer-focused content, SEO improvements, and mobile responsiveness.Direct Webpage Ingestion
Ingest web content with<sm-highlight> tags for targeted extraction.Usage Limits Dashboard
Billing usage and limits now visible directly in the console dashboard.Other
- Allow all CORS origins for easier API integration.
- Single
containerTagin add memory — simpler API for basic use cases. - Improved MCP project handling — better project scoping in the MCP server.
New Consumer App
Complete rewrite of the consumer-facing app — new chat experience with slide-out window, masonry memory grid with infinite scroll, PWA support, and mobile-responsive menu bar.Memory Graph with WebGL
Graph rendering now uses WebGL for smooth visualization of thousands of memory connections. Search highlights relevant nodes with zoom.Chat Rewrite
New chat system with memory-aware conversations, regeneration, copy buttons, and the ability to add memories through chat.Dynamic Node Relations
Memory graph now supportsupdate, extend, and derive relation types. Memories can be inferred from multiple parent documents.PDF Support for Google Drive
Google Drive connector now processes PDF files alongside Docs, Sheets, and Slides.Encrypted Connector Credentials
Google Drive, OneDrive, and Notion client secrets are now encrypted at rest.Bulk Memory Delete
New endpoint for deleting multiple memories at once.Self-Host Support
Initial self-hosting support — run Supermemory on your own infrastructure.Console Migrated to Cloudflare
Console app moved from Vercel to Cloudflare Workers for improved performance and lower latency.Autumn Payments Integration
Billing system integrated with Autumn for subscription management, waitlist early access, and usage tracking.New Developer Dashboard
Redesigned developer dashboard with API key display in code snippets, limits visualization, and MCP installation instructions.Consumer App v0
First version of the consumer app with chat, memory browsing, project management, and profile view. New consumer-oriented landing page.MCP → Agents SDK
MCP server migrated to the Agents SDK architecture for better reliability and project support.New Billing
Revamped billing page with upgrade buttons and plan management.Console Dashboard
First version of the dashboard overview page with memory analytics, container tag distribution charts, and usage metrics.Google Drive Webhooks
Real-time sync — Google Drive changes are automatically detected and processed.Sentry Integration
Error monitoring added across the console and API.Launch-Ready API
Console reached launchable state with login page improvements, auth fixes, and the first version of the new dashboard with React Query.Infinite Chat
Memory Router proxy with automatic context compression for infinite-length conversations with LLMs.Container Tags in Search
Filter search results by container tags for scoped memory retrieval.Google Docs MD Export
Google Drive connector switched from PDF to Markdown export for better content fidelity.API v3
New/v3/ endpoints replacing v2 — cleaner routes, updated memory endpoint, and new update/delete operations.OneDrive Connector
Initial OneDrive integration for syncing files into Supermemory.Connections Architecture
New connection-document relationship model for tracking which connector synced which document.Comprehensive API Documentation
New interactive API references on Mintlify with detailed parameter explanations, response schemas, and bearer auth.Container Tags System
Enhanced organizational grouping for better memory isolation and user-scoped content.Auto Content Type Detection
Automatic processing of PDFs, images, videos, and web content regardless of URL extensions.Google Drive Connector
New endpoints for programmatic Google Drive integration and file syncing.Search Improvements
documentThresholdandchunkThreshold— fine-tune search sensitivity.docIdparameter — search within specific large documents.onlyMatchingChunks— precise result filtering.endUserIdfiltering — scope search to specific users.- Reranking — improved result quality with a reranking step.