Now with Model Context ProtocolRead more

Your agents forget.
Ramorie remembers.

The second brain for developers and their AI agents — one place for tasks, long-term memory, and decisions that every agent can read and write through a single protocol.

npm i -g ramorie · free for solo developers

ramorie · memory

claude-code wrote a memory

Decisionmem_0042

Chose pgvector over a separate vector database — Postgres already holds our data, and it's one less service to operate.

#search#db

Linked to 2 memories · now searchable by every agent

Recalled later

Why did we go with pgvector?
You chose pgvector to keep vectors inside Postgres — one less service to run.mem_0042
Works withClaudeCursorGPTAny MCP client

The problem

Every session starts from zero.

You re-explain your codebase every single session.

Decisions vanish when the context window closes.

Knowledge is trapped in chats no other tool can read.

One system

Four things, one memory.

type=decision · pattern · skill

Memory

Typed, tagged, and linked memories — written once, recalled by every agent you use.

ramorie task add

Tasks

Work items that live next to the context that created them. Agents can read, create, and complete them.

scoped recall

Projects

Every memory and task belongs to a project, so recall stays relevant to the codebase you're in.

MCP · CLI

Agents

Claude, Cursor, or any MCP client connects to the same shared brain — one MCP memory server, a single protocol.

Capture once. Recall everywhere.

01

Agents capture

As you work, your agents write decisions, patterns, and fixes to Ramorie — no manual note-taking.

02

Ramorie organizes

Memories are typed, tagged, linked, and embedded for semantic search.

03

You recall

Ask in plain language from any tool and get answers grounded in your own history.

.mcp.jsonOne config, every agent.
{
  "mcpServers": {
    "ramorie": {
      "command": "ramorie",
      "args": ["mcp", "serve"]
    }
  }
}

Built for agents

Read, write, and reason over one shared brain.

  • Project-scoped recall — agents only see the memory that belongs to the codebase they're working in.

  • Verifiable citations — every answer links back to the exact memories it was grounded in.

  • Idempotent writes — saving the same memory twice converges to one, so agents never create duplicates.

1line to connect
memory outlives the context window
100%answers grounded in your own knowledge

Give your agents a memory that lasts.

Free for solo developers. Connect your first agent in under a minute.