Mem0 — the memory layer for AI agents

Short brief · Kindred Ventures AI-agents panel, SF

For: Jack / Tomales Bay Capital · Date: October 20, 2026Context: CTO Deshraj Yadav on Kindred Ventures (Steve Jang) panel in SF today on AI agents.

What They Do

Mem0 is the memory infrastructure layer for AI agents — a model-agnostic API that lets developers store, retrieve, and evolve user memory across apps, sessions, and LLMs. Agents stop starting from scratch. The pitch: memory becomes as standard as a database.

What Needs to Be True

Assumption	Status	Reason
Agents become the dominant software interface, creating massive memory demand	Looks likely	Every major lab and cloud is betting on this; agentic frameworks already exploding
Big labs keep memory siloed, leaving room for a neutral layer	Unclear	OpenAI/Anthropic have incentive to lock in memory — but enterprise buyers may demand portability
Developers converting to paid at meaningful rates	Unclear	200K devs, 186M API calls/mo — but revenue is undisclosed; OSS-to-paid conversion is hard
AWS partnership signals hyperscaler buy-in (not a one-off)	Looks likely	Named exclusive memory provider for AWS Agent SDK — strong signal; more conversations implied
A ~20-person team can scale the infrastructure before a better-funded competitor closes the gap	At risk	Letta (Felicis), OpenAI and Google are all in the space, so there's little room for execution mistakes (team size per LinkedIn, ~20)

Founders

Taranjeet Singh (CEO) — Ex-Paytm, Khatabook; built one of the first GPT app stores (1M+ users). Seventh company.
Deshraj Yadav (CTO) — Created EvalAI at Georgia Tech; led AI Platform at Tesla Autopilot/FSD.

Sources: mem0.ai/about-us, TechCrunch Oct 2025

Funding

$24M total (announced Oct 2025): $3.9M seed led by Kindred Ventures, $20M Series A led by Basis Set Ventures. Also: YC (S24), Peak XV, GitHub Fund. Angels include Dharmesh Shah (HubSpot), Olivier Pomel (Datadog), Thomas Dohmke (ex-CEO GitHub).

Traction

200K+ developers on cloud platform; 66K+ GitHub stars (company-reported, Oct 2026)
API calls: 35M in Q1 2025 → 186M in Q3 2025 (~30% MoM) (TechCrunch)
13M+ Python package downloads; 486+ research citations
Exclusive memory provider for AWS Agent SDK — biggest third-party validation
Revenue undisclosed. SOC 2, HIPAA, GDPR compliant.

Competitors

Letta (Felicis-backed, UC Berkeley), Zep, Supermemory, and first-party memory from OpenAI/Anthropic/Google. Mem0's edge: neutral, portable, open-source-first across any LLM/framework.

Questions for Deshraj

AWS exclusivity: What did that eval look like, and are other hyperscalers in conversation? Is this a template or a one-off?
Business model: OSS-to-paid conversion rate and revenue split between self-serve vs. enterprise?
Lab threat: If OpenAI ships portable memory, what's your actual moat — is it the data flywheel, the neutral positioning, or something else?

Sources: mem0.ai/about-us · TechCrunch · Kindred Ventures · PR Newswire. Usage/star figures are company-reported or from TechCrunch citing company data — directional only.