The new MemoryRouter: Glacier, sharper, and built like a real product

MemoryRouter just got rebuilt visually from the ground up.

New logo. New palette. New typography. New dashboard feel. New marketing site. New product surface.

The old MemoryRouter looked like what it was at the time: a powerful infrastructure product moving faster than its own brand.

Useful, but uneven.

Technically sharp, visually patched together.

That version is gone.

The new MemoryRouter is built around the Glacier system: cool whites, deep dark surfaces, high-contrast green accents, Fraunces display type, clean Inter body copy, and a sharper visual language that feels like serious AI infrastructure instead of another dev-tool landing page.

This was not a button-color pass.

This was a full company redesign.

Why we redesigned it

MemoryRouter has outgrown its original skin.

It started as a memory layer for agents and OpenClaw workflows. That story still matters, but the product is bigger now.

MemoryRouter gives AI products a private memory layer for every user.

user_id -> Memory Key -> private vault

That means the product is not just a clever proxy. It is infrastructure for AI applications that need continuity across sessions, agents, providers, and product surfaces.

The old design did not carry that weight.

The new one does.

The new visual system

The redesign centers on a colder, more premium infrastructure feel.

The palette moved into Glacier territory: crisp whites, icy gradients, graphite blacks, and a cleaner green accent that still feels like MemoryRouter without looking noisy.

The typography changed too.

Fraunces gives the brand a little edge and confidence. Inter keeps the product readable and practical. Together, they make the site feel more editorial, more premium, and less like every other AI SaaS page with a glowing orb and a pricing card.

The logo was refreshed to match that system.

The new mark is cleaner, colder, and more ownable. It works across the marketing site, dashboard, docs, favicon, social surfaces, and GitHub without feeling like five different brands stapled together.

That consistency matters.

AI infrastructure is trust infrastructure. The product has to feel stable before anyone reads the docs.

The dashboard caught up

The redesign was not limited to the homepage.

The dashboard got the same treatment: cleaner surfaces, sharper contrast, theme-aware components, better spacing, and a more deliberate dark/light system.

Memory products are inherently abstract. Tokens, vaults, usage, retrieval, ingestion, sessions, keys, shards. If the dashboard feels messy, the product feels messy.

The new dashboard makes the system feel calmer.

You can see what is happening. You can understand what each user memory vault is doing. You can manage billing and usage without feeling like you are inside a prototype.

That was the bar.

Not just “make it prettier.”

Make it feel real.

The product story is sharper now

The redesign also forced the messaging to grow up.

The core idea is simple:

Every user in your AI product should have their own private memory layer.

Not chat history.

Not giant transcript stuffing.

Not “summarize the last 10 messages and hope the model cares.”

Actual user-scoped memory.

MemoryRouter supports two integration paths:

Proxy mode

Send model requests through MemoryRouter and it handles retrieval, provider forwarding, and storage automatically.

This is the fastest way to add memory to an existing AI product.

Local inference mode

Keep inference inside your own stack and use MemoryRouter only for memory retrieval and storage.

1. POST /v1/memory/prepare  -> retrieve this user's memory
2. Call your model provider directly
3. POST /v1/memory/ingest   -> store the completed exchange

Your provider routing, streaming, retries, evals, logging, and model gateway stay where they already are.

MemoryRouter handles memory. Your stack handles inference.

That is the product now.

Cleaner design. Clearer architecture. Stronger category.

OpenClaw still matters

OpenClaw remains one of the best ways to experience MemoryRouter because agent memory is such an obvious pain.

Agents forget context. They lose decisions. They rediscover the same project facts. They need continuity across compaction, resets, and restarts.

The mr-memory plugin still solves that.

But the new brand makes the hierarchy clearer:

OpenClaw is a supported integration.

MemoryRouter is the memory layer for AI products.

The new MemoryRouter

This redesign is the product catching up to the ambition.

A better logo.

A colder palette.

A sharper type system.

A real dashboard.

A stronger homepage.

A clearer story.

MemoryRouter now looks like what it is becoming: the private user memory layer for AI products that need continuity.

The infrastructure was already powerful.

Now the brand finally looks the part.