Mem0

Mem0 (pronounced mem-zero) is a memory layer for AI agents and assistants from Embedchain, Inc., which does business as Mem0. It uses an LLM to extract the facts worth keeping from conversations, stores them, and retrieves the most relevant ones when an agent needs them, so preferences and context carry across sessions. Memory is scoped per user, agent or app. Mem0 ships as a managed platform and as open-source software under the Apache 2.0 license, and its GitHub repository has 67k stars. The company raised $24M across a Seed round led by Kindred Ventures and a Series A led by Basis Set Ventures, with participation from Peak XV Partners, GitHub Fund and Y Combinator, announced in October 2025. The managed platform is reached through Python and TypeScript SDKs, a REST API, a CLI and a hosted MCP server. OpenMemory runs locally in Docker, keeps its data on the user's machine and exposes its own MCP server to clients such as Claude, Cursor and Windsurf. The open-source SDK works with LLM providers including OpenAI, Anthropic, Google, AWS Bedrock, Groq, Mistral and local models through Ollama, and with more than 20 vector databases including Qdrant, Pinecone, pgvector and Weaviate. Mem0 publishes integration guides for LangGraph, CrewAI, the OpenAI Agents SDK, Google ADK, the Vercel AI SDK, Mastra, n8n, Zapier and Claude Code, among others. Mem0 says AWS selected it as the exclusive memory provider for the Strands Agents SDK. The platform has four plans, priced in USD and billed monthly. Hobby is free, needs no credit card and includes 10,000 add requests and 1,000 retrieval requests a month, one project and community support. Starter is $19/month with 50,000 add requests, 5,000 retrieval requests and one project. Pro is $249/month with 500,000 add requests, 50,000 retrieval requests, unlimited projects, private Slack support, advanced analytics, graph memory with entity linking and Dream memory consolidation. Enterprise is custom-priced and adds unlimited requests, on-premise deployment, audit logs, custom integrations, SSO and SLA support. Every plan allows unlimited end users, and Mem0 also offers usage-based pricing for teams whose traffic does not fit a tier. Starter at $19/month does not include graph memory, Dream memory consolidation, analytics or more than one project, which start at Pro, and on-premise deployment, SSO and audit logs require Enterprise. Mem0's terms and privacy policy allow content from Hobby plan users to be used to train its AI models and exclude Paid Plan data from training. Mem0 is best for developers and product teams building AI agents or assistants that need per-user memory across sessions, on a managed service or self-hosted. Teams that want to assemble chat history and business data in a temporal knowledge graph for Graph RAG should consider Zep. Developers who want their own work context captured automatically across browsers, IDEs and collaboration tools should consider Pieces for Developers, and teams that want a packaged desktop app capturing their own screen and audio history should consider screenpipe. Current state Q3 2026: Mem0's April 2026 memory algorithm uses single-pass, add-only extraction and retrieval that combines semantic, keyword and entity signals. On its managed platform Mem0 reports 92.5 on LoCoMo and 94.4 on LongMemEval, up from 71.4 and 67.8 for its previous algorithm, at under 7,000 tokens per retrieval call, and says open-source results will be directionally similar but not identical. Temporal reasoning and memory decay followed in May 2026, and on September 16, 2026 Mem0 became available as a native integration on the Vercel Marketplace, billed through the Vercel invoice. Mem0's Trust Center lists a SOC 2 Type 1 report, a HIPAA self-attestation and GDPR, and a BAA is available on Enterprise.