Skip to content

How Lodekit works

You don’t need to understand Lodekit’s insides to use it. But a rough map helps you know where your apps live, where your data goes, and what your agent is actually doing. Here’s the whole picture.

Diagram: how Lodekit's pieces fit together — the manager supervises one runtime process; you use the dashboard in the browser, your agent uses MCP and writes code into the apps folder.Manager (menubar)starts & watchesRuntimeone local Node processDashboardYour appsServicesstorage · tasks · events …GatewaysYouuse apps,tune settingsYour agentClaude, Codex, …MCPscaffold, inspectapps/ folderwrites codeBrowserlodekit.localhost:7777
One local process runs everything; you use the browser, your agent uses MCP and the file system.

Two facts carry most of the mental model.

One process. The engine is a single program on your Mac. It hosts every app you build, plus the dashboard, all served in your browser at http://lodekit.localhost:7777. The manager — the menubar icon — starts the engine, watches over it, and restarts it if you ask.

One folder. Everything you’d ever want to keep lives in the Lodekit root, ~/Documents/Lodekit. Inside it sit three trees, side by side: apps/ holds each app’s source files, data/ holds what each app stores, and gateways/ describes the outside providers your apps can reach. That one folder is the thing you back up.

The engine has two doors, one for each of you.

The dashboard is yours. It shows every app with its live status, and it’s where you tune settings, browse logs, and manage secrets — all in your browser.

Your agent has its own door. Through a separate connection, the agent can create apps, open them for you, and read an app’s logs and data on your behalf — so it can answer “show me all my recipes” or debug an app it built. You’ll set that connection up in Connect your agent.

When your agent creates or changes an app, all it does is write files into apps/. Lodekit’s watcher notices the change, builds the app with its own build tools, and serves the result:

Diagram: the build lifecycle — you save a file, the watcher notices, the shared toolchain builds, and the app is served or the error is shown on the dashboard.You save a filein apps/<app>/Watcher noticesBuildshared toolchainreadyerrorApp servedat /<app-id>/Error shownon the dashboardstate changes stream live to the dashboard (SSE)
Save a file; Lodekit rebuilds and serves. No terminal, no build commands.

You never run a build. Save (or let your agent save), and the app’s card on the dashboard flips to building…, then ready. If something goes wrong, the card shows the error instead — and your agent can read the details and fix it.

Each app keeps what it stores under data/ in its own space, right next to the apps/ tree — so an app’s records, files, and settings are ordinary contents of your Lodekit folder, and backing up that folder preserves all of it. The full layout is in Your Lodekit folder.

Apps don’t reinvent the plumbing. Lodekit provides services every app can use: a database for records, a small memory for quick notes, file storage, events, scheduled tasks, safe doors to outside providers (like the weather forecast), secrets, settings, and logs. Your agent builds with these, which is why even a quick app gets durable storage and a settings page for free. Meet them in the Services overview.

That’s the map. Time to put Lodekit on your machine: Install Lodekit.