Mem0 Secures $24M to Build the Default Memory Layer for AI Agents

Mem0, a San Francisco-based startup developing memory infrastructure for AI agents, announced a $24 million funding round combining a Seed led by Kindred Ventures and a Series A led by Basis Set Ventures.

The round also includes investments from Peak XV Partners, GitHub Fund, and Y Combinator, alongside individual backers such as Scott Belsky, Dharmesh Shah, and CEOs from Datadog, Supabase, PostHog, and Weights & Biases.

The company aims to solve a fundamental gap in AI: persistent memory. Despite advances in large language models, most AI systems still forget user preferences and previous interactions. This limitation forces developers to repeatedly feed context to models, hindering personalization and long-term learning.

Mem0’s infrastructure offers developers a plug-and-play memory layer that integrates with only three lines of code. It stores essential data, removes outdated information, and retrieves relevant context efficiently.

Since launch, Mem0 has achieved 41,000 GitHub stars and 14 million downloads, with API calls rising from 35 million in Q1 to 186 million in Q3 2025.

“Every agentic application needs memory, just as every application needs a database,” said Taranjeet, Co-founder and CEO of Mem0. “We're using this funding to become the default memory layer for AI agents, making LLM memory accessible and reliable for all developers.”

The platform is already being used by thousands of teams, including Fortune 500 companies. Frameworks like CrewAI, Flowise, and Langflow have built-in support for Mem0, and AWS selected it as the exclusive memory provider for its Agent SDK.

The funding comes as competition in AI infrastructure intensifies.

Earlier this month, Anthropic expanded Claude’s enterprise integrations to include long-term memory capabilities, highlighting the growing focus on making AI systems more context-aware and adaptive.