Open Source · MIT Licensed

Drop-in AI workflows
for your NodeNextBunDenoNestNode app

$ npx flowlib-cli initCopy

Visual Workflow Editor

Build complex workflows visually

Connect AI agents, API integrations, and conditional logic into production-ready pipelines — all from a drag-and-drop canvas.

Flowlib flow editor showing a Linear ticket triage workflow with AI agent, switch routing, and integrations

Why Flowlib

An open-source workflow engine embedded in your app

It's a library, not a platform

You npm install it into your Express, NestJS, or Next.js app. It uses your database, your auth, your deployment pipeline. No separate server to maintain, no vendor lock-in, no outgrowing the tool.

See integration guides →

your-app.ts

// Your existing Express app
import { createFlowlibRouter } from '@flowlib/express';

// Mount alongside your existing routes
app.use('/api', yourRouter);
app.use('/workflows', await createFlowlibRouter(...));

// that's it

Not another LangChain wrapper

Flowlib has a custom-built execution engine from the ground up.

Smart branching — inactive paths and their downstream nodes are skipped
Upstream outputs auto-aggregated by node name into downstream inputs
Full JavaScript expressions in config params via QuickJS WASM sandbox

Learn about the execution model →

Incoming data for each node

fetch_users[{ id: 1, name: "Alice" }]

get_config{ env: "prod" }

api_response{ status: 200 }

Full JS in any config field

Process {{ fetch_users.filter(u => u.active).length }} users in {{ get_config.env }}

Native batch processing

Flows automatically pause for OpenAI and Anthropic batch jobs, then resume when results arrive. Batch APIs are 50% cheaper — and no other flow builder handles this natively.

Learn about batch processing →

Fetch data

Transform inputs

AI Model

→ batch submitted, flow pauses→ polling checks batch status→ completes after ~20 min→ flow resumes automatically

Process results

Send notification

Flows as code, synced to GitHub

Define workflows as readable .flow.ts files with the Builder SDK, then sync them to GitHub. Push changes directly, open a pull request on publish, or pull reviewed edits back into the visual editor.

Write typed flows with defineFlow(...) and SDK node helpers
Sync readable .flow.ts files to GitHub with push or pull
Use PR-based publishing for reviewable, auditable workflow changes

Learn about Git sync →

support-triage.flow.ts

import { defineFlow, input, model, output } from '@flowlib/core/sdk';

export default defineFlow({
  name: 'Support triage',
  nodes: [
    input('ticket', {
      variableName: 'ticket',
    }),
    model('classify', {
      credentialId: '{{env.OPENAI_CREDENTIAL}}',
      model: 'gpt-5-mini',
      prompt: 'Classify {{ ticket.title }} ...',
    }),
    output('result', {
      outputName: 'classification',
      outputValue: '{{ classify }}',
    }),
  ],
  edges: [['ticket', 'classify'], ['classify', 'result']],
});