pref0 vs Mem0

Mem0 provides memory for AI agents. pref0 provides preference learning. Both help agents personalize, but they solve different problems in different ways.

pref0Mem0
What it storesStructured preferences with confidence scoresMemories as natural language facts
How it learnsExtracts preferences from corrections and conversationsStores facts from conversations
Confidence over timeYes — preferences compound across sessionsNo — memories are stored as-is
Data modelKey-value preferences with scoresNatural language memory entries
Integration2 REST endpointsSDK with multiple methods
Best forLearning how users want things doneRemembering facts about users

Key differences

Preferences vs. memories

Mem0 stores memories — facts like 'user works at Acme Corp' or 'user has a dog named Max.' pref0 stores preferences — structured learnings like 'prefers TypeScript over JavaScript' with a confidence score of 0.85. These are complementary concerns.

Confidence compounding

pref0's core differentiator is confidence scoring that compounds over time. A preference mentioned once starts at 0.40–0.70. The same preference across multiple sessions reaches 1.0. Mem0 stores memories without confidence scoring.

Correction detection

pref0 specifically watches for corrections — 'use X instead of Y' — as the highest-signal data. This is purpose-built for preference learning. Mem0 treats all conversation content equally as potential memories.

When to use each

Use pref0 when...

  • You want to learn how users prefer things done (tools, style, format)
  • You need confidence scoring that compounds over time
  • Corrections and explicit feedback are your primary signal
  • You want a minimal API (2 endpoints)
  • You're building a coding assistant or task-oriented agent

Use Mem0 when...

  • You need to remember facts about users (name, company, context)
  • You want a broader memory system beyond preferences
  • You need conversation history storage
  • You're building a personal assistant that needs factual recall

Frequently asked questions

Can I use pref0 and Mem0 together?

Yes. They're complementary. Use Mem0 for factual memory (user's name, company, past conversations) and pref0 for preference learning (how they want code written, communication style). Inject both into your agent's system prompt.

Is pref0 a replacement for Mem0?

No. pref0 focuses specifically on preferences with confidence scoring. If you need general memory, Mem0 is the better tool. If you need preference learning, pref0 is purpose-built for it.

Which is easier to integrate?

pref0 has 2 REST endpoints (track and profiles). Mem0 has an SDK with multiple methods. Both are straightforward, but pref0 is simpler if you only need preferences.

Other comparisons

vs Zepvs Custom Memoryvs RAGvs LangChain Memoryvs LangMemvs Lettavs ChatGPT Memoryvs Pineconevs CrewAI Memoryvs Semantic Kernel

Not memory. Preference learning.

Your users are already teaching your agent what they want. pref0 makes sure the lesson sticks.

Get Started FreeView Docs