MemNexus is in gated preview — invite only. Learn more

Give your coding agents
memory that persists

Every new session starts from zero — your codebase conventions, architecture decisions, and debugging history, gone. MemNexus gives Claude Code, Cursor, and Copilot persistent context that survives every restart.

No credit card required · Invite-only preview

Have an invite? Sign in →

~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "memnexus": {
      "command": "mx",
      "args": ["mcp", "serve"]
    }
  }
}
✓ 5-minute setup · Claude Desktop, Claude Code, Cursor, Windsurf, Copilot, Codex
Working Memory · next session
auto-injected
Active Work

Migrating auth from REST to gRPC — schema finalized in PR #47, client stubs left to generate

Key Facts

gRPC schema uses proto3 · buf lint enforced · no streaming RPCs yet

Team uses Zod validation + strict TypeScript, ESLint enforced

18 memories·4 conversations·last 48h

See the Difference

Same task. Completely different experience.

Claude Code
>I need to add webhook notifications to the API
I'd be happy to help you add webhook notifications! To provide the best solution, I have a few questions:
1. What's your current tech stack and framework? 2. How do you handle async jobs currently? 3. Do you have any existing notification patterns I should follow? 4. What's your testing approach?
>Node.js with Express, we use Bull for queues...
Got it. And what retry strategy do you prefer? Do you have an existing notification service I could look at for patterns?
>Check src/jobs/ for our email notifier
Readsrc/jobs/
● No persistent memoryclaude-sonnet-4

Back-and-forth to establish context every session

Claude Code+ MemNexus
>I need to add webhook notifications to the API. Use MX to recall previous work we've done and our established process for creating, testing, and deploying new features.
Memory"notification patterns" "feature development" "testing process"
Found 4 relevant memories from your codebase patterns
I can help with that. Based on your established patterns:
• Your email notifier uses Bull with 3 retries + exponential backoff → src/jobs/email-notifier.ts • New features follow: route → service → job → tests • You use Vitest + MSW for API mocking • Webhooks should include HMAC signatures (per your security decisions)
I'll create the webhook job following your email notifier structure. Starting with the Bull job definition?
● Context loaded from 4 memoriesclaude-sonnet-4

Instant context from your coding history

Sound familiar?

Every developer knows these frustrations

🔄

You've explained your coding style 47 times

🔍

You can't find that conversation where you solved the exact same bug

📋

Your AI gives generic answers, not ones tailored to your workflow

🚀

Every new project starts from zero

Six patterns of AI memory failure — documented with solutions. See the full picture →

Or read: Why AI assistants lose context — and what to do about it

What MemNexus Does

Three things your coding agent can't do without MemNexus

Searchable Decision History

Why you chose PostgreSQL over MongoDB. Which auth library you rejected and why. The debugging session that traced a race condition to your caching layer. Every technical decision is extracted, indexed, and retrievable by meaning — not just keywords.

Learns Your Team's Conventions

Your strict TypeScript config. Your Zod validation patterns. Your preference for async/await over .then() chains. MemNexus extracts these from your conversations and applies them automatically — so your agent stops suggesting approaches your team already rejected.

Git-Aware Memory (CommitContext)

Every git commit captures what changed — MemNexus captures why. The architectural reasoning, the tradeoffs considered, the bugs discovered. Six months from now, git blame tells you who changed a line. MemNexus tells you the full story.

Works however you work

🖥️CLICommand-line interface
🔌MCPCursor, Windsurf, Claude Desktop
📦SDKPython, TypeScript, JS
🔗APIRESTful API

See how it works → · Building with the SDK

Why Not Just Use Built-In Memory?

ChatGPT Memory, Cursor context, and Copilot workspace indexing are useful starting points -- but they keep your knowledge trapped in one tool. Learn the architectural difference.

Cross-Tool

Built-in memory

Locked to one tool. ChatGPT memory only works in ChatGPT. Cursor context only works in Cursor.

MemNexus

Works across Claude Code, Cursor, Copilot, Windsurf, and any dev tool that supports MCP or REST APIs. Switch tools without losing context.

Persistent

Built-in memory

Memory resets between sessions, disappears when you switch machines, and gets wiped by tool updates.

MemNexus

Survives session resets, tool switches, and machine changes. Your knowledge graph lives in the cloud, accessible from anywhere.

Learns Over Time

Built-in memory

A flat list of facts with no relationships. Can't connect that your Redis caching decision relates to your scaling strategy.

MemNexus

Builds a knowledge graph that links decisions, patterns, and context. Understands relationships between your memories, not just individual facts.

How It Works

From install to persistent memory in under 5 minutes

1

Install and Connect

npm install the CLI, run mx setup, and it auto-detects your tools — Claude Code, Cursor, Windsurf, Copilot. One command writes the MCP config. Your workflow doesn't change.

2

Code With Your Agent

Discuss architecture, debug a failing test, choose between libraries. MemNexus extracts the decisions and reasoning from your conversations — why you chose Zod over io-ts, what caused that flaky integration test, which caching strategy you ruled out.

3

Context Carries Forward

Next session, your agent already knows your stack, your conventions, and what you tried last week. It retrieves relevant context before you type your first message — across tools, projects, and team members.

terminal

Setup guides: Claude Code · Cursor · Windsurf · Full walkthrough →

Common questions

Everything you need to know before getting started.

Simple, transparent pricing

Full learning capabilities in Free tier

Free

Experience the magic

$0
  • 200 memories/month
  • Full memory & learning
  • 60-day retention
Most Popular

Pro

Unlimited for power users

$10/month
  • Unlimited memories
  • Forever retention
  • Advanced search

Enterprise

For teams that ship

Custom
  • Team collaboration
  • SSO/SAML
  • Self-hosting option

Get started in under 5 minutes

Sign up, install the CLI, and run mx setup. Works with Claude Code, Cursor, GitHub Copilot, Windsurf, and more.

Already have an account? Sign in →