memctl

mem/ctl

Shared, persistent memory for AI coding agents.
One project brain across machines and IDEs.

Website · Docs · Changelog

---

What is mem/ctl?

mem/ctl is a cloud MCP server that gives AI coding agents shared, persistent memory scoped to projects and organizations. Claude Code, Cursor, Windsurf, Cline, Roo Code, Codex, OpenCode, and any MCP-compatible agent can read and write to the same project context.

Your team shares one brain. Every agent knows the architecture, conventions, decisions, and constraints without repeating yourself.

Quick start

npx memctl auth
npx memctl init

The init wizard detects your IDE and writes the MCP config automatically. Or configure manually:

{
  "mcpServers": {
    "memctl": {
      "command": "npx",
      "args": ["-y", "memctl@latest"],
      "env": {
        "MEMCTL_ORG": "<org-slug>",
        "MEMCTL_PROJECT": "<project-slug>"
      }
    }
  }
}

Agent support

Agent	Config file	Init shortcut	Generate
Claude Code	.mcp.json	memctl init --claude	memctl generate --claude
Cursor	.cursor/mcp.json	memctl init --cursor	memctl generate --cursor
Windsurf	~/.codeium/windsurf/mcp_config.json	memctl init --windsurf	memctl generate --windsurf
Cline	.vscode/mcp.json	memctl init --cline	memctl generate --cline
Roo Code	.vscode/mcp.json	memctl init --roo	memctl generate --roo
Codex	codex.json	memctl init --codex	memctl generate --codex
OpenCode	opencode.json	memctl init --opencode	memctl generate --opencode

All agents, all at once:

npx memctl init --all

MCP tools

mem/ctl exposes 11 tools to connected agents:

Tool	What it does
memory	Store, get, search, list, update, delete memories with dedup and tagging
memory_advanced	Batch get, find similar, diff, merge, and rename keys
memory_lifecycle	Archive, restore, history, bulk tag/untag, and capacity checks
context	Bootstrap project context, delta sync, smart retrieval, rules evaluation
context_config	Manage typed project guidance (coding style, architecture, constraints)
session	Session lifecycle, conflict claims, and rate status
activity	Memos, activity log, impact analysis
branch	Branch-scoped context and implementation plans
repo	Repository structure analysis
org	Organization info and project listing
import_export	Bulk import/export memories as JSON

Plus 7 resources and 3 prompts for richer agent integration.

How it works

1. Agent connects to memctl via MCP (stdio transport)
2. On startup, the agent bootstraps project context (conventions, architecture, constraints)
3. During the session, the agent reads and writes memories as needed
4. All agents on the team share the same project memory
5. Session tracking prevents write conflicts across concurrent agents

Features