Comparison · Vodou vs Mem0 · checked Sep 17, 2026

The Mem0 alternative for memory you use yourself.

The model changes. Your memory doesn't.

Most people searching for a Mem0 alternative want one of two things. If you are building an AI product and need memory for its users — Python and TypeScript SDKs, a hosted API that scales, graph memory, SOC 2 — stay with Mem0 or compare it with Zep and Letta. If you want memory for yourself, the job the retired OpenMemory extension and MCP server did, Vodou is that alternative: an app on your computer that stores memory in a local SQLite file, types it into ChatGPT and Claude, and injects it into Claude Code and Cursor. Vodou needs a free account and is not an SDK.

Download Vodou See the comparison

Vodou vs Mem0Personal memory, not an SDKWorks with local modelsFree Vodou account requiredmacOS · Windows · Linux

TL;DR

Mem0 is a memory layer developers build into their own AI apps — Apache-2.0 open source with 65,476 GitHub stars, plus a hosted Mem0 Platform; Vodou is a memory and harness one person installs and uses directly.
Vodou stores memory in a SQLite file on your computer and searches it on-device; the Mem0 Platform, its hosted MCP server and its Claude Code plugin keep memory in Mem0's cloud, while Mem0 open source runs on infrastructure you host, with OpenAI models by default.
Vodou captures and types memory into 22 AI chat sites through the Vodou Bridge Chrome extension — a job Mem0 stopped doing when it archived the OpenMemory Chrome extension on March 23, 2026.
Vodou costs $10 a month on your own model keys after a 60-day free trial; Mem0 open source is free, and the Mem0 Platform is free up to 10,000 adds a month, then $19 or $249 a month.

Definition

What is Mem0?

Mem0 is a memory layer for AI agents and apps: an Apache-2.0 library and self-hostable server for Python and TypeScript, plus the hosted Mem0 Platform with an API, an MCP server and plugins for Claude Code, Cursor and Codex. Developers call add and search so their product remembers each user. Vodou, compared here, is personal memory instead.

Mem0 runs three ways: as a library inside your code (local Qdrant by default), as a self-hosted server with a dashboard, per-user API keys and an audit log (Postgres with pgvector), or as the managed Platform. Open source defaults to OpenAI's gpt-5-mini and text-embedding-3-small, and can be configured for Ollama or LM Studio. Graph memory, memory decay, webhooks and structured export are Platform features. Mem0 says 150,000+ developers use it and lists SOC 2 Type 1 and HIPAA.

OpenMemory was Mem0's product for individuals. Its Chrome extension was archived on March 23, 2026, and OpenMemory was removed from Mem0's repository on July 29, 2026. The OpenMemory page now sends people to Mem0's Claude Code plugin and its hosted MCP server, whose documentation says: “Nothing runs on your machine: the server is hosted by Mem0.”

The problem

Why do people look for a Mem0 alternative?

Searches for Mem0 alternatives show four groups: developers who want memory running locally without Docker, a vector database and an OpenAI key; former OpenMemory users whose Chrome extension and self-hosted MCP server were retired in 2026; teams comparing graph or temporal memory in Zep, Letta or Cognee; and people who wanted a personal memory, not an API. Vodou fits the first, second and fourth groups.

OpenMemory was retired

The Chrome extension that captured ChatGPT, Claude and Perplexity chats is archived, and the self-hosted MCP server left the repository. What remains stores memory on the Mem0 Platform.

Self-hosting Mem0 is infrastructure

Local Mem0 means running Qdrant or Postgres, choosing embedding and extraction models, and wiring an MCP server yourself — which is why the “local” search results are community wrappers.

An API isn't a memory you use

Mem0 remembers what your code sends it. If you want memory that follows you across your own AI tools, you need capture, inject and a place to fix mistakes. For a product's graph memory, Zep or Letta fit better than Vodou.

Capabilities

How is Vodou different from Mem0?

Mem0 gives developers building blocks; Vodou is a finished app for one person. Vodou captures memory itself from 22 AI sites, Claude Code, Cursor and messaging channels, stores it in SQLite with full-text, vector and reranked search on your computer, shows why each memory was used, and lets your own agents read it through an MCP host and an OpenAI-compatible API on 127.0.0.1.

Capture is built in

Mem0 stores what your code sends to add(). Vodou collects memory itself: the Vodou Bridge Chrome extension on 22 AI sites, hooks in Claude Code, Cursor and Codex, every messaging channel you connect, and importers for ChatGPT, Claude, Obsidian, OpenClaw, Hermes and Letta.

Stored and searched on your machine, nothing to host

Memory is a SQLite file with full-text search, on-device embeddings and a cross-encoder reranker — no Qdrant, Postgres or Docker to run. Extraction uses the model you pick; choose a local model and nothing leaves the computer.

Memory on every prompt in Claude Code

Mem0's Claude Code plugin recalls up to five memories on a session's first prompt, then searches when asked. Vodou's hooks inject matching memory at session start and on every prompt, and extract new facts at session end.

Your own agents can use it too

Any MCP client attaches to Vodou's MCP host with its own token and a vault that limits what it reads; scripts can call an OpenAI-compatible API on 127.0.0.1 with memory behind it. Single-user, on your machine — not a multi-tenant API.

See why a memory was used, and fix it

Every retrieval shows its source, date and trust weight. Correct, reject, pin or undo a memory in the Vodou Console, and the fix applies in the browser, the editor and every channel at once.

Memory that runs tasks

What Vodou remembers feeds unlimited skills, MCP servers and schedules, and answers in Slack, Telegram or WhatsApp — so the context that recalls a decision can also act on it.

Comparison

Vodou vs Mem0: how do they compare?

Mem0 wins for developers: Python and TypeScript SDKs, memory for unlimited end users, hosted scale, graph memory on Pro, SOC 2 and HIPAA, a free tier, Apache-2.0 code and 65,476 GitHub stars. Vodou wins for one person's own AI use: capture from 22 AI sites and Claude Code, local storage with no database to run, memory on every Claude Code prompt, messaging channels, and memory that runs skills and tools.

→Scroll the table sideways to see 2 columnsall 3 columns

CAPABILITY	VODOU personal memory + harness	MEM0 PLATFORM hosted API, MCP, plugins	MEM0 OPEN SOURCE self-hosted, Apache-2.0
Built for	One person using AI tools	Developers adding memory to apps and agents	Developers who run their own infrastructure
How memory gets in	Captured for you: 22 AI sites, Claude Code / Cursor / Codex hooks, channels, importers	Your code calls add; plugins for Claude Code, Cursor, Codex; hosted MCP server	Your code calls add()
Where memory lives	A SQLite file on your computer	Mem0's cloud	Your infrastructure — local Qdrant (library) or Postgres + pgvector (server)
Search	Full-text + on-device vectors + cross-encoder reranker	Multi-signal retrieval; graph memory on Pro and Enterprise	Vector search; graph memory is Platform-only
Model that extracts memories	The one you choose — Claude, GPT, or local Ollama, LM Studio, llama.cpp	Run by the Platform	gpt-5-mini + text-embedding-3-small by default (OpenAI key); Ollama and LM Studio configurable
SDKs and API	No SDK; MCP host and OpenAI-compatible API on 127.0.0.1	Python and TypeScript SDKs, REST API, hosted MCP server	Python and TypeScript SDKs; self-hosted REST server
Many end users	No — single user	Unlimited end users on every plan	Yes — per-user API keys on the self-hosted server
Memory from ChatGPT, Claude and other chat sites	22 sites via Vodou Bridge, typed into the box visibly	No — OpenMemory Chrome extension archived March 2026	No
Claude Code	Hooks: session start and every prompt; extraction at session end	Plugin: up to 5 memories on the first prompt, then search; sends batches every 5 exchanges	Community MCP servers (e.g. mem0-mcp-selfhosted)
Memory shared across a team	No	Shared project memory per repository (Claude Code plugin)	If you build it
Messaging apps	Unlimited channels (Slack, Telegram, Discord, WhatsApp…)	Through your code, n8n or Zapier	Through your code
Memory can run tasks	Yes — unlimited skills, MCP servers and schedules	No — memory layer	No — memory layer
Compliance	None claimed; runs on your own machine	SOC 2 Type 1, HIPAA, GDPR ready	Depends on your deployment
Export	Portable Vodou pack (vodou-core mem export)	Schema-driven export jobs through the API	It's your database
Code you can read	Gateway, extension, skills, MCP servers Apache-2.0; engine proprietary	Service closed; core library Apache-2.0	Apache-2.0 — 65,476 GitHub stars
Account	Yes, free (licenses the engine)	Yes — Mem0 API key	No
Price	$10/mo on your own model keys after a 60-day free trial	Free (10k adds, 1k retrievals/mo) · $19/mo · $249/mo · Enterprise	Free; you pay for hosting and model calls

Sources checked (2026-09-17); Mem0 columns are as of that date.

Mem0 Platform — mem0.ai · pricing · Platform vs open source · Mem0 MCP · memory export · Claude Code plugin (v0.3.1) · OpenMemory page
Mem0 open source — github.com/mem0ai/mem0 · open-source overview · Ollama LLM config · mem0-mcp-selfhosted (community)
OpenMemory — Chrome extension repo (archived 2026-03-23) · PR #6530, OpenMemory removed 2026-07-29 · mem0ai/openmemory

Star count from the GitHub API on 2026-09-17 (mem0.ai shows an older figure). Vodou column: 22 AI sites, 4 hooks and the importers checked against the product on 2026-09-17.

Which to pick

When should you choose Mem0 instead of Vodou?

Choose Mem0 if you are building a product: you need memory for many end users, Python or TypeScript SDKs, a hosted API with SOC 2 and HIPAA, graph memory, webhooks, or Apache-2.0 code you can self-host without an account. Choose Vodou if the memory is for you — across ChatGPT, Claude, Claude Code, Cursor and messaging apps — and you would rather install one app than run a vector database and an extraction model.

Mem0 fits better if…

You are adding memory to an app or agent that serves other people.
You need Python or TypeScript SDKs and a hosted API that scales.
You need SOC 2, HIPAA, SSO or on-prem deployment.
You want graph memory, memory decay or webhooks.
You need every line Apache-2.0 and no account to self-host.
Your team should share project memory in Claude Code.

Vodou fits better if…

The memory is yours, across ChatGPT, Claude and 20 other AI sites.
You want memory on every prompt in Claude Code and Cursor.
You used OpenMemory and want a maintained extension and local MCP host.
You want local storage without running Qdrant, Postgres or Docker.
You want the same memory in Slack or Telegram, running skills and tools.

How it works

How do you move from Mem0 to Vodou?

Vodou has no Mem0 importer, so memory moves as Markdown: install Vodou, export your Mem0 memories to .md files, import that folder with vodou-core mem import, then switch Claude Code, Cursor and your MCP clients from Mem0 to Vodou so only one memory injects.

STEP 1
Install Vodou.
Download Vodou for macOS, Windows or Linux and sign in with a free account. The installer wires Claude Code and Cursor hooks; add Vodou Bridge from the Chrome Web Store if you also want memory from ChatGPT, Claude and the other 20 AI sites — the job the OpenMemory extension used to do.
STEP 2
Export your Mem0 memories.
On the Mem0 Platform, create a memory export job (POST /v1/exports/) or page through get_all for each user_id. On Mem0 open source, call get_all on your own instance. Write each memory's text as a bullet in a Markdown file — one file per user or project keeps them organised.
STEP 3
Import the folder with vodou-core mem import.
Vodou has no Mem0 importer, but it reads Markdown: run vodou-core mem import obsidian <folder> --dry-run to preview, then without --dry-run. The importer reads any folder of .md files, sanitizes them, and labels each fact as imported so it ranks below what you tell Vodou directly. vodou-core mem import-undo rolls an import back.
STEP 4
Switch your tools over.
Uninstall the Mem0 Claude Code plugin (claude plugin uninstall mem0@mem0-plugins) so two memory hooks don't both inject, replace the Mem0 MCP entry in Claude Desktop, Cursor or VS Code with Vodou's MCP host, then search the Vodou Console for facts you rely on and correct anything wrong.

FAQ

Frequently asked questions about Mem0 alternatives

What are the best Mem0 alternatives?

For developers adding memory to an AI product, the alternatives most often compared with Mem0 are Zep (temporal knowledge graph), Letta (agents that manage their own memory), Cognee and Supermemory. For a person who wants memory across their own AI tools — ChatGPT, Claude, Claude Code, Cursor and Slack — Vodou is the alternative: an app on your computer rather than an API you build on.

Is there an open source alternative to Mem0?

Mem0 itself is open source under Apache-2.0, and so are Letta, Zep's Graphiti and Cognee; Supermemory is MIT. Vodou is open-core: its gateway, Chrome extension, skills and MCP servers are Apache-2.0 in the VodouAI/OS repository, while the Rust engine that stores and searches memory is proprietary. If fully open-source code is a requirement, Vodou does not meet it.

Is there a free alternative to Mem0?

Mem0 is already free to self-host, and its hosted Hobby plan is free for 10,000 memory adds and 1,000 retrievals a month. Letta, Graphiti and Cognee are free to self-host too. Vodou is not free: it is free for 60 days with no credit card, then $10 a month on your own model keys, and it needs a free Vodou account that licenses the engine.

What is a local alternative to Mem0?

Vodou runs locally without a database to operate: memory is a SQLite file on your computer with full-text search, on-device embeddings and a reranker, and extraction can use Ollama, LM Studio or the bundled llama.cpp so nothing leaves the machine. Mem0 open source can also run locally, but you host Qdrant or Postgres and configure a local model instead of its OpenAI defaults.

What happened to OpenMemory?

Mem0 archived the OpenMemory Chrome extension on March 23, 2026, and removed OpenMemory from its main repository on July 29, 2026. The mem0ai/openmemory repository is now a tool for moving coding sessions between Claude Code, Codex and OpenCode, and mem0.ai/openmemory points to Mem0's Claude Code plugin and its hosted MCP server. Vodou Bridge and Vodou's MCP host cover what the extension and self-hosted server did.

What is the difference between Mem0 and OpenMemory?

Mem0 is the memory layer — the open-source library, the self-hosted server and the hosted Platform API. OpenMemory was Mem0's personal-memory product built on it: a Chrome extension and a self-hosted MCP server for tools like Cursor and Claude. Both of those are retired; the OpenMemory name now leads to Mem0's Claude Code plugin and hosted MCP server, which store memory on the Mem0 Platform.

Can I use Vodou as a memory layer for my own agent?

For an agent you run yourself, yes: connect it to Vodou's MCP host with its own token and vault, or call the OpenAI-compatible API on 127.0.0.1, and it reads and writes the same local memory. For a product with many users, no — Vodou is single-user, has no Python or TypeScript SDK and no hosted API. That is the job Mem0, Zep and Letta are built for.

Does Vodou send my memory to the cloud?

Vodou stores and searches memory on your computer, and the free account that licenses the engine never holds your memory. What leaves your machine is what goes to the model you picked — your prompts, and the extraction step that turns captured chats into facts. Choose Ollama, LM Studio or the bundled llama.cpp and that stays local as well.