Quickstart

Connect one agent to a Lore server and save your first memory. Choose Loremem Cloud for the fastest start, or run a server on your own machine with Docker.

Option A: Loremem CloudCopy Anchor Link

You need Node.js 20 or later and the agent runtimes you want to connect.

1. Sign in at https://app.loremem.com/login with GitHub or Google. Your first sign-in creates a personal organization and a default project.

2. Use the project switcher at the top of the console to pick the project. Then open the Install page.

3. Enter a Token note that says where you will use the token, such as laptop, and select Create token.

4. Select Copy command. The command includes your token. The token and command are shown once, so copy them before you close the dialog.

5. Run the command in a terminal:

   Copy Text

   npx @loremem/cli install --base-url https://api.loremem.com --api-token lm_your_token

   The installer uses the server and token from the command, and connects the runtimes it detects on this machine. A runtime that isn't installed is reported as skipped. To choose runtimes yourself, add --channels, for example --channels claudecode,codex.

6. Restart each agent runtime. Hooks and MCP configuration load when a runtime starts.

Option B: Self-hostedCopy Anchor Link

You need the following. The installer supports macOS and Linux.

• Node.js 20 or later, to run npx @loremem/cli.
• Docker with Docker Compose.
• An OpenAI-compatible embedding endpoint (base URL, API key, and model).
• A chat model for View LLM. Lore accepts OpenAI-compatible Chat, OpenAI Responses, or Anthropic Messages.

1. Run the installer:

   Copy Text

   npx @loremem/cli

   In a terminal, choose Deploy the server on this machine with Docker. For a non-interactive run, use this command instead:

   Copy Text

   npx @loremem/cli install -y --docker

   The installer downloads docker-compose.yml into ~/.lore/docker, writes a .env with a random database password next to it, starts the stack, waits until it is healthy, writes ~/.lore/config.json, and connects the runtimes it detects.

2. Open http://127.0.0.1:18901/setup and finish the four setup steps:

   1. Embedding: the base URL, API key, and model for your OpenAI-compatible /v1/embeddings endpoint.
   2. View LLM: the provider, base URL, API key, and model. Lore uses it to write memory views and to run Dream.
   3. Global boot memories: core://agent, core://soul, and preferences://user. Select Skip to save the default text for an empty boot node and continue.
   4. Channel agent memories: core://agent/<runtime> for each runtime you connect.

   Setup saves settings in the database, not in .env. You can change them later at /settings.

3. Restart each agent runtime.

Verify it worksCopy Anchor Link

Check these four things in order. The first two confirm that memory is saved and recalled. The last two confirm that the console sees the activity.

1. Boot. Start a new session in the project. Ask the agent what startup memory it received. It should describe the boot nodes you set up, such as core://agent and preferences://user.
2. Write. Ask the agent to save a real project fact, such as the example in the next section. Then open Memory (in Loremem Cloud) or /memory (self-hosted) and find the new node. A new memory graph has little to recall, so start with a fact you will reuse.
3. Recall. Send a prompt with the same words, for example "How should I test refund webhooks in checkout_api?" Then ask the agent to quote the <recall> block it received for that prompt. If the new memory appears as a line, recall works.
4. Console activity. In Loremem Cloud, open Overview to see Queries, Shown, and Used, plus recent queries. On a self-hosted server, open Analytics (/recall/drilldown). The Queries count rises after each prompt.

Write your first memoryCopy Anchor Link

Ask the agent to save a durable fact. For example:

Save this as a long-term memory for the checkout_api project. Create the parent node project://checkout_api first. Then create project://checkout_api/webhooks with priority 1, a disclosure for changing refund or payment webhooks, and the glossary terms refund webhook, idempotency key, and stripe. Content: refund webhooks are idempotent by event.id, and tests must replay the same event twice and assert a single refund row.

The saved memory has this shape:

URI:        project://checkout_api/webhooks
Priority:   1
Disclosure: When changing refund or payment webhooks in checkout-api
Glossary:   refund webhook, idempotency key, stripe
Content:    Refund webhooks are idempotent by event.id. Tests must replay the same event twice and assert a single refund row.

• URI is the stable address. Use snake_case segments and keep dates out of the path. A child path needs its parent node first, so project://checkout_api must exist with its own content, disclosure, and glossary.
• Priority 1 marks a key fact. Lore allows up to 15 nodes at priority 1.
• Disclosure describes one situation in which the memory should surface. If you find yourself writing "A or B", write two memories instead.
• Glossary lists the words people actually type, including product names and error names.
• Content is the fact itself, in plain text.

For each field in more detail, see Writing good memories.

Next stepsCopy Anchor Link

• Move CLAUDE.md and AGENTS.md into Lore: move shared rules out of local files.
• Connect agents: add the other runtimes on this machine.
• How Lore works: see what each agent receives and records.
• The web console: inspect and fix memories in the browser.
• Server configuration: change settings on a self-hosted server.