What is Mem0, and is it open source or a paid platform?

Short answer

Mem0 is a memory layer for AI agents and apps: it stores facts from conversations outside the model, so an agent remembers across sessions. It comes in two forms. Mem0 Open Source is an Apache 2.0 library or Docker stack you run on your own infrastructure. Mem0 Platform is the hosted version, which as of November 2026 has a Free plan with usage limits and a Custom plan on request.

What Mem0 does

A language model forgets everything between sessions. Mem0 is a separate layer your app or agent calls to save and recall facts. In code it comes down to two calls:

  1. Add. You pass a conversation or a message. Mem0 uses an LLM to pull out facts worth keeping (“prefers invoices in PDF”, “works in Amsterdam”) and stores them, tied to an identifier such as a user_id.
  2. Search. Before the model answers, you search with the question and that identifier. Mem0 returns the most relevant memories, which you put in the prompt.

The README describes the current algorithm, introduced in April 2026, as single-pass and add-only: one LLM call per add, and nothing is overwritten, so memories accumulate. Search combines semantic similarity, BM25 keyword matching and entity matching.

Mem0 is a building block for developers. It does not plug into ChatGPT or Claude for you; your code, or an MCP server, decides when to add and when to search.

Open source or Platform

Mem0 Open Source Mem0 Platform (hosted)
What Library (pip install mem0ai, npm install mem0ai) or Docker stack Hosted API with an API key
License Apache 2.0 Commercial service
Data On your infrastructure With Mem0, or on-prem on Custom
Default models gpt-5-mini and text-embedding-3-small from OpenAI, configurable Mem0 runs model calls for the first 30 days, then your own LLM key (Free)
Extras What you build Custom adds graph memory, user profiles, memory consolidation, analytics, SSO, audit logs, SLA

Pricing, as of November 2026

Plan Price Adds Retrievals
Free No fee 10,000 a month for 30 days, then 50,000 a month with your own LLM key 1,000 a month for 30 days, then 10,000 a month with your own LLM key
Custom On request Set to your volume Set to your volume

Usage past the Free limits is billed by usage, with a monthly cap you set.

The common misunderstanding: the benchmark numbers

Mem0 publishes strong scores on long-memory benchmarks such as LoCoMo and LongMemEval, and a research paper (arXiv 2504.19413). Two things to read alongside them. The README says the scores reflect Mem0’s managed platform, and that open source users should expect “directionally similar gains but not identical numbers”. And these benchmarks test whether a system recalls facts from long conversations with one user. That is a different question from whether a team’s knowledge reaches every colleague’s AI tool.

What this means for a team

Mem0’s model is memory per identifier: per user, per agent, per session. That fits an app where each customer should have their own memory. For a team it raises practical questions you have to answer yourself: who is the “user” when five colleagues work on the same client, who may read which memories, and how a fact one person’s agent stored reaches another person’s ChatGPT or Copilot. Mem0 gives you the store and the search; the sharing rules are your build.

If you are weighing Mem0 against similar tools, see Mem0 vs Zep vs Supermemory vs Letta and Mem0 alternatives for team memory. For the MCP route, read does the Mem0 MCP server give a whole team shared memory.

Where Bombos fits

Mem0 is a memory library you build on; Bombos is a plugin a team installs. It gives every AI tool the team uses (Claude, Claude Code, ChatGPT, Copilot, Grok, Codex, Cursor, your own agents) one shared memory with team knowledge and rules the AI learned about how you work, and it reads what already sits in mail, documents and repos, with a trail of what each answer rests on. It runs in our cloud or on your own server, and one person can try it first with their own agents.

Related questions

Is Mem0 open source?

Yes. The mem0ai/mem0 repository on GitHub is licensed under Apache 2.0. You install it with pip install mem0ai or npm install mem0ai, or run the self-hosted server with Docker.

How much does Mem0 cost?

The hosted Free plan includes 10,000 adds and 1,000 retrievals a month for the first 30 days, then 50,000 adds and 10,000 retrievals a month with your own LLM key. The Custom plan has no listed price.

Can I self-host Mem0?

Yes, with the open source version, as a library or a Docker stack, with your data on your own infrastructure. The hosted Custom plan also lists on-prem deployment.

Which LLM does open source Mem0 use by default?

According to the README, gpt-5-mini from OpenAI for extraction and text-embedding-3-small for embeddings. You can configure other providers.

Does Mem0 work with Claude Code or Cursor?

Mem0's documentation has guides for Claude Code, Cursor, Codex and other coding agents, and it offers an MCP server. See our page on the Mem0 MCP server for what that does and does not share.

Sources

  1. mem0ai/mem0 (GitHub README)
  2. Mem0 documentation: introduction
  3. Mem0 pricing
The first half hour · free

What does your AI forget?

  1. 01You tell us which AI you use and where it keeps starting over.
  2. 02We show what a shared memory changes there, on your own examples.
  3. 03You keep the plan, with or without Bombos.
Piet Baudoin
Piet BaudoinYou talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe
We reply the same working day.