Skip to content

What is Flowcraft?

Flowcraft is a lightweight, unopinionated, and progressively scalable runtime for executing declarative workflows defined as directed acyclic graphs (DAGs). It is designed to reliably orchestrate complex business processes, data pipelines, ETL workflows, or AI agent orchestrations with a focus on simplicity, portability, and extensibility.

Core Philosophy

Unlike heavy platforms like Temporal or Airflow, or domain-specific libraries like LangChain, Flowcraft is a foundational engine that does one thing exceptionally well: execute a graph of functions defined as data. It provides a flexible, type-safe API to define workflows, execute them with resilience, and scale from in-memory scripts to distributed systems without changing the core business logic.

Flowcraft offers two primary ways to author workflows: the battle-tested Fluent API for manual graph construction and the new Compiler for writing imperative TypeScript code that automatically generates declarative blueprints.

Key Features

Use Cases

Flowcraft is versatile for various workflow scenarios.

AI Agents

Build intelligent agents that process data, make decisions, and interact with users.

  • Example: Research Agent (see Research Example)
  • Features: Conditional branching, LLM integration, human-in-the-loop.

ETL Pipelines

Extract, transform, and load data efficiently.

  • Example: Parallel Workflow (see Translation Example)
  • Features: Batch processing, parallel execution, error handling.

Business Process Automation

Automate routine business tasks like approvals and notifications.

Distributed Execution

Run workflows across multiple machines or services.

  • Example: Distributed Workflow (see Distributed Example)
  • Features: Adapters for queues, persistence.

Choose the right pattern for your needs!

Released under the MIT License