Model Context Protocol Standard · MCP 2025-03-26

Persistent Long-Term Memory for AI Agents

Sub-2ms persistent memory server. Your agents remember architecture decisions, user preferences, and project rules across sessions. Seamlessly integrates with Claude, Cursor, Codex, and any MCP client.

Start for Free Quickstart Guide

mcp.mmwhub.tech — live

$ curl -I https://mcp.mmwhub.tech/mcp

HTTP/2 401 Unauthorized (OAuth 2.0 required)
server: FastMCP 2.0
via: Caddy · Let's Encrypt TLS
www-authenticate: Bearer resource_metadata="https://mcp.mmwhub.tech/.well-known/oauth-protected-resource"
latency: <2 ms

< 2 ms

Ultra-low Latency

In-memory SQLite WAL on NVMe. Zero vector-DB overhead for instant agent recalls.

Zero-Leak

Privacy First

Isolated tenant encryption. Stores atomic facts and references, never raw source files.

~800 B

Memory Footprint

Extracted micro-facts preserve tokens and context windows. 95% cheaper than prompt re-feeding.

100%

Open Protocol

Native Model Context Protocol compatibility across Claude, Cursor, Codex, Windsurf & Aider.

Use Cases

Built for Serious Engineering

A single memory protocol across dev environments, automated pipelines, and agent teams.

Continuous Architectural Decisions

Stop repeating architectural decisions every time you start a new chat. Agents recall prior ADRs, API contracts, and verified bugfixes.

Recall prior incident fixes, design choices, and root causes.
Retrieve strict coding guidelines without LLM hallucinations.
Cryptographically isolated workspaces per repository or team.

// Query: Auth service contract v2

$ mmw search "jwt refresh rotation rule"

[FACT-8812] VERIFIED

Enforce strict token rotation with 15-minute TTL. No plaintext tokens in logs (ADR-2026-03).

Stateful Context for AI Agents

Static AGENTS.md files bloat context windows and rot. MMW provides dynamic, structured recall — your agents remember architectural rules, past bugfixes, and code guidelines across sessions.

200 tokens injected on demand instead of 20,000 bloated in prompts.
Agents recall previous decisions, never repeating past mistakes.
Continuous cross-session continuity between IDE, terminal, and background tasks.

// Rule Update

$ mmw remember "Never commit secrets or touch /app/auth without approval"

[RULE SAVED]

Rule propagated instantly across all connected IDEs and agent runners.

Shared Team Workspaces

Architecture decisions, forbidden libraries, and verified bugfixes are never lost when switching tools. Claude Code, Cursor, Windsurf, Codex — one single memory.

Shared engineering history, postmortems, and verified architectural decisions.
Automatic decision capture without manual copy-pasting between chats.
Massive token cost savings: precise micro-facts retrieved on demand.

// Codex → MMW → Cursor

sync time: 1.1s

[MEMORY SYNCED]

Session resilience fix verified. Monkey-patch rolled back. Hardened rule committed to memory.

Hardened Security & Isolation

Enterprise-grade memory fabric built for production environments. End-to-end cryptographic isolation, role-based access control, and zero third-party vector-DB dependencies.

Dedicated tenant encryption and immutable audit trails for every memory call.
Hardware-backed key encryption and cryptographic integrity verification.
OAuth 2.0 PKCE standard — API tokens are never leaked to logs or disk.

// Security & Compliance Audit

$ mmw audit --compliance=soc2

[AUDIT PASS: COMPLIANT]

Plaintext tokens excluded from logs. OAuth 2.0 PKCE enforced.

Integrations

Works in Your Existing Tools

One single endpoint for all MCP clients. Set up in under 60 seconds.

Claude.ai & Desktop

OAuth 2.0 PKCE one-click authorization. Native support in browser and desktop apps.

https://mcp.mmwhub.tech/mcp

Cursor IDE

Add in Settings → Features → MCP Servers. Transport: SSE.

Type: SSE | URL: https://mcp.mmwhub.tech/mcp

Codex CLI & Aider

Direct connection via environment variable or config.toml.

MMW_ENDPOINT=https://mcp.mmwhub.tech/mcp

Pricing

Simple, Transparent Pricing

Worldwide Visa, Mastercard & local payment methods supported via secure billing.

Free

For exploring the protocol and personal experiments.

3-day trial

Up to 250 memory records
Up to 60 calls / min
MCP Gateway included
No credit card required

Start for Free

POPULAR

Starter

For active developers, freelancers, and small teams.

per month

Up to 5,000 memory records
Up to 120 calls / min
Multi-device synchronization
Priority response <2ms

Get Started

Pro

For full-time engineers and heavy AI agent workflows.

$35

per month

Up to 100,000 memory records
Up to 120 calls / min
Priority response <2ms
Dedicated engineering support

Get Started

Enterprise

For large-scale agent swarms and enterprise deployments.

$117

per month

Up to 1,000,000 memory records
Up to 300 calls / min
Custom NDA & Data Privacy Agreement
99.95% SLA & Dedicated Support Engineer

Get Started