MemoryLake vs MemMesh: A Memory Passport vs. an Installable Brain

MemoryLake and MemMesh start from a similar conviction: your memory shouldn't be trapped inside one vendor's chat app. MemoryLake calls its version a 'Memory Passport' — a portable, encrypted memory that follows you across ChatGPT, Claude, Gemini, and any model over an API. We built MeshKey on the same instinct, so we come at this comparison with genuine respect for the shared premise. Details below are from MemoryLake's public site and docs as of July 2026; some of its scale claims are self-reported and we treat them as such.

What MemoryLake does well

• Portable 'Memory Passport' — context that follows the user across models, not locked to one app.
• Six structured memory types — background, fact, event, conversation, reflection, skill — through a unified API.
• Explicit provenance — it advertises knowing where each piece of knowledge came from, when, and who accessed it.
• Git-like versioning and conflict detection with audit trails, plus GDPR under its compliance list.

That's a serious feature set, and the portability and provenance overlap with things we care about too. As with Zep, provenance is not on our list of differentiators — MemoryLake has it.

One thing to clear up: the 'Router'

MemoryLake ships a component called the 'MemoryLake Router.' It's easy to assume that's a marketplace, but it isn't — it's an LLM model-router that picks which model provider to call and handles retries and token quotas. It routes across models, not across a catalog of installable brains. Worth naming, because the words are close but the thing is different.

Difference 1: prediction

MemoryLake extracts, structures, and recalls memory; it doesn't forecast future events with calibrated confidence. A MemMesh brain does — declare a target, get a scored prediction from the subject's history, and watch it recalibrate against real outcomes. That's the axis none of the portable-memory products play on.

Difference 2: a marketplace of installable brains

A Memory Passport carries your personal context between models — one person, many tools. A brain is the next unit up: a domain-scoped, publishable memory that anyone can install, with per-brain provenance. MemMesh's Mesh Router is a marketplace and gateway for those brains — owned, purchased, or pinned — so expertise itself becomes portable, not just your own history. MemoryLake has no equivalent.

Difference 3: deployment

MemoryLake is a proprietary cloud SaaS with no self-hosted or on-premise tier we could find. A MemMesh brain runs on-device and private by default, with self-host and VPC options for teams that need them. If data residency is a hard requirement, that's a real fork in the road.

When to pick MemoryLake

If you want a polished, portable personal-memory product that spans every major model out of the box, with structured memory types and provenance, and a hosted cloud suits you — MemoryLake is a strong option and a well-built one. Pick MemMesh when you also need forecasting, an on-device footprint, or a brain you can publish and sell rather than just carry.

MemoryLake makes your memory portable. MemMesh makes expertise portable — a brain any model can install, that also tells you what's next.

We'll keep this current as MemoryLake ships. If it adds calibrated prediction or a true brain marketplace, we'll update this here.