Skip to content

getting-started.md

Latest commit

 

History

History
116 lines (77 loc) · 5.22 KB

getting-started.md

File metadata and controls

116 lines (77 loc) · 5.22 KB
Raw

Getting Started

Memory Engine is permanent memory for AI agents. Store, search, and organize knowledge that persists across conversations.

Were you invited to a shared space? Head to Joining a Space for a teammate-focused walkthrough — logging in, selecting the space, and searching what's already there.

Install

curl -fsSL https://install.memory.build | sh

This installs the me binary to ~/.local/bin. Make sure it's on your PATH.

Sign up and log in

me login

This opens your browser to sign in via GitHub or Google (an OAuth 2.1 auth-code + PKCE flow over a 127.0.0.1 loopback redirect) and stores your credentials. On a host with a system keychain they're stored there; otherwise they fall back to ~/.config/me/credentials.yaml (mode 0600).

On a headless host with no local browser (an agent harness in a sandbox, a remote SSH session, a container), use me login --device instead: the CLI prints a short URL and code to approve on any device (the OAuth 2.0 device authorization grant), yielding a rolling 7-day session token. See me login.

If you belong to more than one space, pick the active one (it's carried as the X-Me-Space on every request):

me space list
me space use <slug-or-name>

me login <space> selects it in one step, and me whoami shows your identity and active space.

If your CLI is older than the server (or vice versa), me login will tell you and bail out before sending you to the browser. You can run the same check explicitly:

me version

Store your first memory

me memory create "PostgreSQL 18 supports native UUIDv7 generation." \
  --tree share/notes/postgres \
  --name uuidv7 \
  --meta '{"topic": "database"}'

A --tree is required. Put memories the rest of your space should see under share/*, and personal ones under ~/* (your private home). The optional --name gives the memory a filename-like slug (unique within its tree) so you can later address it by path -- me get share/notes/postgres/uuidv7. See Core Concepts.

Search

# Hybrid search (meaning + keywords)
me memory search "UUID generation in Postgres"

# Keyword search
me memory search --fulltext "UUIDv7"

# Pure semantic search (by meaning only)
me memory search --semantic "database-generated identifiers"

Browse the tree

me memory tree

Browse in the web UI

For a richer, visual experience there's a web UI with a tree explorer, hybrid / advanced search, a rendered Markdown viewer, and an editor for content + metadata.

  • Hosted (no install): open api.memory.build and sign in with GitHub or Google — the same account you'd use for me login. This is the quickest way in if you don't want to touch the CLI.
  • Local: run me serve to start the same UI against your CLI session on http://127.0.0.1:3000 (or the next free port). See me serve.

Connect to AI tools

Register Memory Engine with your AI coding tools:

me opencode install
me codex install
me gemini install

For a guided, per-project setup that goes further than install — backfilling existing sessions, enabling automatic capture going forward, and adding a memory pointer to CLAUDE.md/AGENTS.md — run me project init once per project. It's harness-agnostic: it detects whichever of Claude Code/OpenCode/Codex you actually have installed and have sessions for, and only offers the steps that apply:

me project init              # guided per-project setup

For Claude Code, me claude install installs the one user-scoped Memory Engine plugin (hooks + slash commands + MCP) — run it once, it applies to every project:

me claude install            # full plugin (once, user scope)
me claude install --mcp-only # or just the MCP server

This drives Claude Code's native plugin flow for you (claude plugin marketplace add + claude plugin install), then persists your resolved server + active space as global defaults and asks whether to capture your Claude Code sessions as memories. Capture is off by default; opt in and new sessions (plus a one-time backfill of your existing ones) are captured privately under ~/projects/<repo> — sharing with a team is a separate, per-project choice via .me/config.yaml. Afterwards, restart Claude Code (or run /plugin) to load the hooks and slash commands; re-run /plugin → memory-engine → Configure to adjust options, or re-run me claude install to change the capture answer.

After installation, your AI agent has access to memory tools -- create, search, get, update, delete, and more.

See MCP Integration for details.

What's next