A powerful new way to quickly build internal tools, automations, admin panels, and much more.

LAUNCH

Simplify persistance

There's no need to connect a database (although you can!) because Membrane programs run in a durable JavaScript runtime.

That's right, we transparently persist memory changes so that you don't worry about loading and storing state.

Any JavaScript object can persist indefinitely because Membrane's execution timeout is effectively

Keep long-lived values in state.
import { state } from "membrane";
state.notes = state.notes ?? []
export async function saveNote({ args }) {
  state.notes.push(args.note);
}

Auditable

The most time-consuming part of bugfixing can often be reproducing the issue in the first place. We didn't like that, so we fixed it!

In Membrane, every side-effect is first written to the program's write-ahead log and then processed. HTTP requests and responses, source code changes, queries, etc are all program side-effects.

This log gives you clear understanding plus the ability to replay history so that reproducing bugs is a walk in the park.

Membrane logs are not just text output; they are the source of truth!

The Graph

In Membrane, each program defines a graph API to expose its data and functionality to other programs. You can think of the graph as objects that can be referenced by other programs, allowing you to easily compose functionality.

In this video we showcase the graph in a quick tour of the Membrane IDE.

Showcase
FEATURES
EXAMPLES
example twitter bot
Uses the Twitter Driver to post a random word definition generated via OpenAI.
Use the OpenAI driver to generate a tweet
Tweet!
import { nodes, root, state } from "membrane";

export async function tweet() {
  const result: any = await nodes.openai.models
    .one({ id: "o3-mini" })
    .completeChat({
      messages: [
        {
          role: "system",
          content: `You're a Ai Dictionary that share lesser-known
          words and fascinating vocabulary facts,
          Return the text with the following format:
          Word: <english word>
          Definition: <short definition>
          Pronunciation: <pronunciation>
          Example sentence: <example sentence>`,
        },
        {
          role: "user",
          content: "Generate a random word",
        },
      ],
      temperature: 1
    });

  await nodes.twitter.tweet({
    text: `${result.content} #membrane`
  });
}
Capabilities
FEATURES
EXAMPLES
Uses the Twitter Driver to post a random word definition generated via OpenAI.
Use the OpenAI driver to generate a tweet
Tweet!
import { nodes, root, state } from "membrane";

export async function tweet() {
  const result: any = await nodes.openai.models
    .one({ id: "o3-mini" })
    .completeChat({
      messages: [
        {
          role: "system",
          content: `You're a Ai Dictionary that share lesser-known
          words and fascinating vocabulary facts,
          Return the text with the following format:
          Word: <english word>
          Definition: <short definition>
          Pronunciation: <pronunciation>
          Example sentence: <example sentence>`,
        },
        {
          role: "user",
          content: "Generate a random word",
        },
      ],
      temperature: 1
    });

  await nodes.twitter.tweet({
    text: `${result.content} #membrane`
  });
}
The Write-Ahead Blog

If you're curious about Membrane's architecture, Juan joined the devtools.fm podcast to talk through the nuts and bolts.



Enter your email for occasional updates