← Memlin/docs/plugin-install-guide
Installation

Plugin install guide

Claude Code

8 steps~2 minmacOS

Install the CLI, open the right project, then add, sign in, bind, and verify — one command at a time.

Do you have the Claude Code CLI?
Great — jumping straight to opening your project and the plugin commands.
Where you type these
The /memlin-* steps below are slash commands typed at the Claude Code prompt — not in a terminal. Launch Claude Code from inside the repo you want bound and /memlin-add-project binds that directory. (On Cursor, Codex, Windsurf, or Antigravity the same actions are the memlin CLI run in a terminal — memlin add-project/memlin-add-project.)
1

Open a terminal in your project

Open a terminal inside the repo you want Memlin bound to — cd into it, or right-click the folder and choose "Open in Terminal", or use your editor's built-in terminal. /memlin-add-project binds whatever directory you launch from — the wrong folder is the #1 cause of project: none.

cd ~/code/your-repo
2

Launch Claude Code

Start Claude Code from inside that directory.

claude
Now inside Claude Code
3

Add the Memlin marketplace

/plugin marketplace add memlin-ai/memlin-claude-plugin
No git? (common on Windows)

The /plugin marketplace add command above runs git clone under the hood, so it fails without git (Command 'git' not found). To install with zero git, exit Claude Code, paste this one line in a terminal — it downloads the plugin over HTTPS and registers it locally.

macOS / Linux · bash
curl -fsSL https://memlin.ai/install-claude.sh | bash

Then restart Claude Code and skip ahead to Sign in — the installer already did the marketplace add + install steps for you.

4

Install the plugin

/plugin install memlin@memlin-ai
5

Reload plugins

Claude Code prompts you to run this — it activates the new plugin's hooks, slash commands, and MCP server.

/reload-plugins
6

Sign in

A browser opens — approve the device code.

/memlin-login
7

Bind this repo (required)

Run once per repo. This is the step most people miss. Open Claude Code at a workspace root holding several sibling git repos (e.g. ~/Repos/Drip/{drip-api,web,mobile}) and Memlin auto-resolves to the org project that owns them — no add-projectneeded; you only need it for a single standalone repo that isn't attached yet, or a non-git folder.

/memlin-add-project
8

Health check

Confirms auth, API reach, and a resolved project.

/memlin-doctor
9

Verify

Healthy shows project: <name> — not none.

/memlin-status
10

Keep it updated (two steps)

When a new version ships, update in two steps — refresh the marketplace mirror first, then update the plugin. /plugin update on its own checks your localcopy of the marketplace, which can lag the remote, so it may report “already latest” while a newer version is actually published. Restart Claude Code afterward to load the new version.

/plugin marketplace update memlin-ai /plugin update memlin@memlin-ai
Don't skip add-project
Sign-in alone leaves the resolver unbound — /memlin-status will show auth: valid, project: none and every prompt returns an empty bundle. Run /memlin-add-project once per repo.
Optional · hand memory to Memlin

This host's built-in auto-memory (MEMORY.md) keeps its own private notes between sessions in ~/.claude/projects/<repo>/memory/. Now that Memlin captures your learnings and serves only the relevant ones — ranked, governed, and shared across your team — that native store is a redundant, machine-bound silo that quietly grows every session. We recommend handing memory to Memlin.

Before you change it:sign in and bind the repo first, so Memlin's capture is already live. Turning the native memory off only stops this host from keeping its own copy — Memlin keeps capturing, so you lose nothing.

To turn it off: Easiest — let Memlin do it: run memlin ingest-native-memory to pull your existing notes in, then memlin manage-memory to hand the toggle to Memlin (reversible anytime with memlin manage-memory --revert). Prefer to do it by hand? Type /memory and toggle auto-memory off, set "autoMemoryEnabled": false in ~/.claude/settings.json, or launch with CLAUDE_CODE_DISABLE_AUTO_MEMORY=1.

Never locked in: everything Memlin stores is exportable anytime — one click in the web app under Settings → Export memory (Markdown or JSON), or from any terminal with memlin pull (/memlin-pull inside Claude Code). The files are readable without Memlin and yours to back up or migrate — Memlin never deletes, it archives.

Vendor settings change — steps verified 2026-06-29.

Disable or uninstall
Feature coverage across agents

Memlin targets the same surface on every adapter. A cell shows whether support is full, partial, not yet shipped, or limited by the host platform (n/a — a constraint of that client, not a Memlin gap). Hover any cell for detail.

AdapterMCP toolsHooksCommandsRules / skillLocal syncScribe
Claude Codefullfullfullfullfullfull
Cursorfullfullfullfullfullfull
Codex CLIfullfulln/afullfullfull
Antigravityfullfullfullfullfullfull
VS Codefulln/an/afullfullpartial
Devin (Windsurf)fullfulln/afullfullfull
Claude Desktopfulln/apartialn/an/apartial
Other MCP clientsfulln/apartialn/an/apartial
Memlin
SAVE TIME. SAVE TOKENS. KEEP MOVING.
Edit this page