MCP-native · for AI agents

One mind.
Many AIs.

Collective intelligence for AI.

What you tell one AI becomes context for another AI and your team. Decisions, learnings, and next steps stay in one Hive so the next AI can use them.

Memory ⇌ Tasks · one feedback loopAIs + humans, one shared layerMCP-native · open standard
Works with
Claude CodeCodex CLIOpenClawClaude DesktopChatGPTCursorVS CodeGeneric MCPClaude CodeCodex CLIOpenClawClaude DesktopChatGPTCursorVS CodeGeneric MCP

Shared hive

Share a hive. Every AI remembers together.

Invite your team to a hive, and each member's AI reads across it. Shared decisions and next steps carry from one person's AI to another's.

Invite

Invite your team in

Add people to a hive by invite. Membership stays inside the circle you choose, so shared memory never leaks past it.

Read-only

Read across hives

A shared hive stays read-only. Each member's AI reads across their own hives and the shared one, and shows where every memory came from.

Fork

Fork to build on it

To build on a shared memory, fork it into your own hive. You extend your copy and link back, without touching the original.

Inside Melxis

Memory that stays useful for work.

Melxis is not a note dump. Mels and Tasks stay connected, so memory can become the next move and completed work can strengthen memory.

Hive

Change the AI. Keep the premise.

Claude, Codex, OpenClaw, and any MCP client can return to the same Hive. Start anywhere and come back to the same context.

Mel

Make context reusable.

A Mel keeps one judgment, reason, or premise as a reusable unit of memory. The next AI can use it as context immediately.

Mel · design decisionsaved

# use hosted MCP endpoint

reason: keeps every client on the same shape

→ linked to related premise and task

Map

Open one memory. See what is nearby.

Related Mels stay linked. From one memory, you can trace the nearby reasons, decisions, and next work.

Tasks

Memory and work improve each other.

Tasks keep active intent alive. Finished work can return reasons and learnings to Mels, so the next AI inherits the loop.

How it works

Change the AI. Keep the thread.

  1. 01

    An AI saves in the moment

    The AI you are talking to records decisions and learnings through MCP. It is not a filing system you clean up later.

  2. 02

    Melxis keeps the relationships

    Mels connect to related memories. Tasks track work that is still moving. Memory and action do not drift apart.

  3. 03

    Another AI continues

    Any MCP-capable AI client can search the same Hive and resume the Task. Change tools without losing the premise.

Connect

Works where you already work.

Melxis is MCP-native, built on the open protocol for connecting AI to external tools. Same shape from every client. No proprietary SDK. No vendor lock-in.

FullGuidedTools

Choose your AI client

Claude Code

01Add
Terminal
claude plugin marketplace add melxis-com/toolkit
claude plugin install melxis@melxis-com-toolkit
02Connect
Terminal
claude
Claude Code
/mcp
Authenticate Melxis
03Collect

As you work with agents, Mels and next steps collect.

Every AI you connect reads the same memory. Share a Hive, and whoever you share it with reads it too.

Human Console

A human-readable console for the shared knowledge your AIs use. Inspect Hives, Mels, Tasks, Graphs, invites, and plans.

Open console

Pricing

Start free. Grow when you need to.

Start by collecting Mels. When you need more care, Apis keeps your daily memory in shape.

Collect
Polish

Free

Try it out

$0
  • 2 Hives · 50 Mels
  • Graph included
  • MCP access
Get started

Lite

Collect more

$3/mo
  • 5 Hives · 300 Mels
  • Graph included
  • MCP access
Get started
Recommended

Basic

Daily memory upkeep

$10/mo
  • 5 Hives · 1,000 Mels
  • Graph included
  • MCP access
  • Apis (Basic): daily memory care
Get started

Pro and Team plans are coming soon.

Start by connecting an AI.

Choose the AI you already use and connect Melxis. Your shared knowledge starts there.

Choose an AI to connect