mem0-api

Mem0 API & SDK

Mem0 is a managed memory layer for LLM agents. It extracts facts from conversations, deduplicates, and exposes them via semantic search — giving agents persistent, cross-session context without you managing a vector DB.

Two products, one mental model:

• Mem0 Platform (managed, recommended) — hosted API + dashboard, sub-50ms retrieval
• Mem0 OSS (self-hosted) — run your own vector store, LLM, and embedder stack

Quick Product Routing

User imports `from mem0 import MemoryClient`      → Platform
User imports `from mem0 import Memory`             → OSS (Python)
User imports `import { Memory } from "mem0ai/oss"` → OSS (Node)
User imports `import MemoryClient from "mem0ai"`   → Platform (TS/JS)
No package installed yet                           → Recommend Platform first

---

Install

# Python
pip install mem0ai

# Node / TypeScript
npm install mem0ai

# CLI
pip install mem0-cli    # Python CLI
npm install -g @mem0/cli  # Node CLI

API Key: Get from app.mem0.ai/dashboard/api-keys

Auth header: Authorization: Token <YOUR_API_KEY>

---

Platform SDK — Core CRUD (TypeScript / Python)

TypeScript (primary for this project)

import MemoryClient from "mem0ai";

const client = new MemoryClient({ apiKey: process.env.MEM0_API_KEY! });

// ADD — async, returns event_id
const result = await client.add(
  [{ role: "user", content: "I moved to Austin last month." }],
  { user_id: "alice" }
);
// result.event_id → poll GET /v1/event/{id}/ to confirm SUCCEEDED

// SEARCH — V3 hybrid (semantic + BM25 + entity). Filters REQUIRED in V3.
const hits = await client.search("Where does Alice live?", {
  filters: { user_id: "alice" },
  top_k: 10,
});

// GET ALL
const all = await client.getAll({ user_id: "alice" });

// GET ONE
const mem = await client.get("<memory_id>");

// UPDATE
await client.update("<memory_id>", { text: "Alice lives in Austin, TX" });

// DELETE ONE
await client.delete("<memory_id>");

// DELETE ALL for a scope
await client.deleteAll({ user_id: "alice" });