Get started with Memora Browser.

Install

Memora Browser needs Node.js 20 or later. The npm package is being prepared for release; once it is published you can run it directly with npx:

Until then, run it from the source folder:

cd memora-browser
npm install
npm start -- --port 9222

Start the server

With no command, Memora Browser starts an automation server. By default it serves the Chrome DevTools Protocol on ws://127.0.0.1:9222 with stealth mode on. Use --protocol webdriver to serve Selenium instead.

Fetch a page

fetch loads one URL, runs its scripts and prints it in the format you choose. See Fetch & dump for examples.

MCP server

Add this to your client’s MCP config (for example mcp.json in Cursor or claude_desktop_config.json). The full tool list is on the MCP server page.

{
  "mcpServers": {
    "memora-browser": {
      "command": "npx",
      "args": ["-y", "memora-browser", "--mcp"]
    }
  }
}

Connect a client

Start the server, then connect instead of launching a browser. More detail on the CDP & WebDriver page.

import { chromium } from 'playwright';

const browser = await chromium.connectOverCDP('http://127.0.0.1:9222');
const context = browser.contexts()[0];
const page = context.pages()[0] || await context.newPage();
await page.goto('https://example.com');

import puppeteer from 'puppeteer-core';

const browser = await puppeteer.connect({ browserURL: 'http://127.0.0.1:9222' });
const page = await browser.newPage();
await page.goto('https://example.com');

from selenium import webdriver

driver = webdriver.Remote("http://127.0.0.1:9222", options=webdriver.ChromeOptions())
driver.get("https://example.com")

Use it as a library

Import the engine directly to load pages, read the action map and act on it without running a server.

import { MemoraBrowser } from 'memora-browser';

const browser = new MemoraBrowser({ stealth: true });
const page = browser.newPage();
await page.goto('https://app.example.com/login');

const snapshot = page.getSemanticSnapshot();
// snapshot.actionMapText, snapshot.markdown, snapshot.tokenEstimate

page.type('@1', 'agent@example.com');
const branch = page.fork();   // copy of the page and its session
page.click('@3');

Memora CDP and HTTP API

Alongside the standard CDP domains (Page, Runtime, DOM, Network, Target, Input), the server adds a Memora domain and two HTTP endpoints:

• Memora.getSemanticSnapshot — action map, actions, Markdown, token estimate and interactive count
• Memora.act — run an action by ID, for example { actionId: "@1", type: "type", text: "…" }
• Memora.getMarkdown — the current page as Markdown
• GET /memora/snapshot?id=<pageId> — the snapshot as JSON over HTTP
• GET /memora/goto?url=<url> — navigate and return the snapshot in one request

Network options

Both the server and fetch accept these options.