@memloom/server

The HTTP layer for memloom, a local-first memory engine for AI agents. A thin Hono app over @memloom/core that validates requests, maps them to engine calls, and serves the viewer.

You probably want the memloom CLI instead. It bundles this server, the store, the viewer and the daemon into one command: memloom init.

npm install @memloom/server

Usage

import { createServer } from "@memloom/server";

const app = createServer(memloom);   // a Memloom instance from @memloom/core

Returns a Hono app you can serve however you like. The daemon in the memloom package uses @hono/node-server.

What it exposes

Memories, recall, context documents, the entity graph and schema, conflicts, reconciliation, and the assistant. Streaming endpoints report progress as NDJSON, so long operations like indexing or a reconcile sweep never hold a request open in silence.

The full route list is the API reference.

Access

Binds to localhost, and both an Origin and a Host check run on every request, so a page on the open web cannot drive a daemon running on someone's laptop.

Apache-2.0. Built by Versuno.