Mem0
A memory layer for AI agents that lets them remember users and past interactions across sessions.
Overview
Mem0 is an open-source (Apache-2.0) memory layer for AI agents. It gives them persistent memory across sessions by extracting and storing facts so they stop forgetting users and context. The San Francisco company raised a $24M Series A in October 2025 led by Basis Set Ventures and reports tens of millions of downloads. In 2025 AWS selected Mem0 as the memory provider for its Strands Agent SDK. It is available as a self-hosted open-source library or a managed cloud platform with an API.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 16 Sep 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with Mem0
- Give an agent persistent memory so it stops forgetting users between sessions
- Extract and store facts automatically rather than re-sending a transcript
- Use it free under Apache-2.0, or on the hosted service
- Adopt the memory provider AWS selected for its Strands Agent SDK
- Judge adoption on 60,000+ GitHub stars and tens of millions of downloads
- Decide your retention and deletion policy first, because persistent memory of users is personal data
What this is for: Adding long-term, personalized memory to AI agents and assistants so they recall users and prior context.
Who this is for: Developers building agents, chatbots, or copilots that need persistence beyond a single conversation.
Availability: Free open-source library plus a paid managed cloud with API; used as the memory provider for AWS's Strands Agent SDK.
What people say
Mem0 has strong adoption signal — tens of millions of downloads, a $24M Series A, and selection as the memory provider for AWS's Strands Agent SDK — and it's a common default when developers want to bolt persistent memory onto an existing agent.
But agent memory is exactly the space where hype outruns evidence, and Mem0's benchmark claims have drawn direct scrutiny. Zep published a rebuttal arguing Mem0's LoCoMo comparison misconfigured competitors (Zep says its corrected score was 75.1% versus the 66% Mem0 reported for it). More damningly, critics note Mem0's own numbers show a plain full-context baseline (~73%) beating Mem0's best (~68%) — evidence the benchmark doesn't actually stress long-term memory, since LoCoMo conversations fit inside modern context windows and lack knowledge-update tests. An arXiv analysis reported Mem0 OSS retrieval precision as low as ~0.06, severe enough that memory could leave a model worse off than none. A broader complaint: write-path cost (often 80%+ of agent runtime) goes unreported across the field.
It's popular and easy to adopt, but treat the state-of-the-art memory claims with skepticism. Benchmark it on your own data before trusting its retrieval quality.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Mem0's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Mem0 unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Mem0 directly →
Spotted something out of date? Suggest an update →
Alternatives to Mem0
More in Agents & Automation
