Choosing your setup

Status: ✅ Current · Last reviewed: 2026-06-18

There's no single "tier" to pick. You choose which AI you use and then how deep you want the integration. Start with the lightest thing that works and add layers only if you need them.


The fastest path: connect a cloud AI (no install)

If you use Claude Web, Claude Desktop, or the ChatGPT app, just add Memeri as an MCP connector — no install, no console, no tunnel. Your AI can read and write your Memeri data (jobs, updates, decisions, intelligence), and every tool call shows on your Project timeline.

This is all most people need to get value. See the per-client pages under Connect Your AI.


Want local file edits + terminal access? Two ways

A cloud AI connected via MCP can't touch your local files. To let an AI run commands and edit files on your machine, pick one:

  • Claude Code (the CLI) — connect it in three layers (below). The richest setup.
  • The Tunnel — a small CLI that gives any MCP client (ChatGPT, Codex, Cursor) local terminal + file access. See Installing the Tunnel.

Or skip local entirely and run agents in the browser Terminal (Claude Code / Codex inside the platform — backed by the local Console).


Claude Code — three independent layers

Each layer is independently useful; do them in order and stop wherever you like:

Layer What it adds Needed?
1. MCP config Read / write your Memeri data Required — this alone connects you
2. User-level telemetry Your local file edits, bash, and git commits appear in the Control Room Recommended
3. Plugin Slash commands + cognition / auto-grounding injection Optional — currently in beta

Full steps on the Connecting Claude Code page.


Quick guide

You want… Do this
Just connect, no install A cloud AI as an MCP connector (Claude Web / Desktop, ChatGPT)
Local file + terminal access Claude Code, or the Tunnel + any MCP client
Run agents in the browser The platform Terminal
The richest integration Claude Code (all three layers)

You can mix these — e.g. ChatGPT as a connector and Claude Code locally. They share the same project memory.