Autoflow

Autoflow is a visual workflow automation platform that lets you design, observe, and replay complex cross-service workflows as if you were stepping through a time machine for your product operations. Teams use Autoflow to orchestrate multi-step flows across AI providers, webhooks, forms, and messaging platforms, get deterministic executions via Inngest, and inspect every run end-to-end from a single, unified graph.

Autoflow is built on Next.js, Node.js, Inngest, tRPC, and Prisma, with a ReactFlow-based editor for the visual layer.

---

Architecture (high-level)

flowchart LR
  subgraph Client["Client / Browser"]
    UI[Next.js App Router UI<br/>React 19 + Tailwind]
    Editor[Visual Workflow Editor<br/>ReactFlow + custom nodes]
    AuthUI[Auth Screens<br/>Login / Signup]
  end

  subgraph Edge["Next.js Edge & API Routes"]
    TRPC[API Gateway<br/>tRPC routers]
    AuthAPI[Better-auth Routes]
    Webhooks[External Webhooks<br/>Google Forms, HTTP]
    InngestHandler[Inngest HTTP Handler]
  end

  subgraph AppCore["Autoflow App Core (Node.js)"]
    WorkflowsSvc[Workflows Service<br/>CRUD + Graph Persistence]
    ExecutionsSvc[Executions Service<br/>History + Status]
    CredentialsSvc[Credentials Service<br/>Provider Secrets]
    Engine[Execution Engine<br/>Topological Sort + Node Dispatch]
    Channels[Real-time Channels<br/>Status Streaming via Inngest]
  end

  subgraph Integrations["External Integrations"]
    OpenAI[(OpenAI)]
    Anthropic[(Anthropic)]
    Gemini[(Gemini)]
    Discord[(Discord Webhooks)]
    HttpAny[(HTTP Targets)]
    GoogleForms[(Google Forms)]
  end

  subgraph Data["Persistence & Infra"]
    DB[(PostgreSQL via Prisma)]
    Prisma[Prisma Client]
    Sentry[(Sentry)]
  end

  UI --> Editor
  UI --> AuthUI

  Editor --> TRPC
  AuthUI --> AuthAPI
  Webhooks --> TRPC

  TRPC --> WorkflowsSvc
  TRPC --> ExecutionsSvc
  TRPC --> CredentialsSvc

  InngestHandler --> Engine
  Engine --> Channels

  Engine -->|Node Executors| OpenAI
  Engine --> Anthropic
  Engine --> Gemini
  Engine --> Discord
  Engine --> HttpAny
  GoogleForms --> Webhooks --> Engine

  WorkflowsSvc --> Prisma --> DB
  ExecutionsSvc --> Prisma --> DB
  CredentialsSvc --> Prisma --> DB

  AppCore --> Sentry
  Edge --> Sentry
  Client --> Sentry

Loading

---

Key capabilities

• Visual workflow editor: Drag-and-drop nodes (triggers, HTTP calls, AI model invocations, messaging) onto a canvas to build complex flows without writing orchestration code.
• Deterministic executions with replay: Workflows are executed via Inngest with full history, making it easy to replay, debug, and audit executions.
• Multi-provider AI orchestration: Built-in nodes for OpenAI, Anthropic, and Gemini, wired to user-scoped credentials with a consistent execution contract.
• Trigger ecosystem: Manual triggers, Google Form triggers, and webhooks that can fan into the same execution engine.
• Typed API surface: tRPC-based API with end-to-end TypeScript types between client and server.
• Production-ready observability: Sentry integrated across client, edge, and server paths for traces and errors.

---

Getting started (development)

# install dependencies
bun install

# run dev server
bun run dev

Then open http://localhost:3000 and:

• Create a workflow under Workflows
• Add triggers and execution nodes in the Editor
• Run the workflow and inspect executions under Executions

---

Tech stack overview

• Frontend / App Shell: Next.js App Router, React 19, Tailwind CSS, ReactFlow
• Backend / Orchestration: Node.js, Inngest functions, tRPC, Prisma ORM, PostgreSQL
• Auth: better-auth (server + client), credential-based provider access
• Observability: Sentry (server, edge, client), structured error reporting
• Tooling: ESLint, Prettier, TypeScript, Bun/npm for scripts