Open Source — MIT licensed. Star and contribute on github.com/pandastack-io/pandaflow
Open Source · MIT

PandaFlow

Build AI workflows visually.

Open source visual workflow builder for AI agents. Connect LLMs, databases, APIs, and code execution nodes on a drag-and-drop canvas. Powered by Firecracker isolated microVM execution.

160+ Nodes
MIT Licensed
Self-hostable
Firecracker microVMs
PandaFlow — AI Workflow Canvas

Trigger

HTTP Webhook

LLM Node

Gemini 2.5 Pro

Code Execution

Python · Firecracker VM

Output

JSON Response

▸ Workflow triggered via webhook...
✓ LLM responded in 1.2s
✓ Python executed in Firecracker VM
✅ Result returned: 200 OK

Everything you need to build agent workflows

PandaFlow gives you the primitives to wire together AI agents, code execution, and data flows — without writing orchestration boilerplate.

Drag-and-Drop Canvas

Build complex agent pipelines visually on a React Flow canvas. Connect nodes with a click — no YAML, no code required.

160+ Built-in Nodes

LLMs, databases, APIs, webhooks, HTTP requests, data transformations, conditional branching, and more — out of the box.

Isolated Code Execution

Run Python, Node.js, Go, and Bash in isolated Firecracker microVMs. No shared runtimes, no cold-start delays.

Real-time Monitoring

Watch your workflow execute node-by-node with live logs streamed via Server-Sent Events. Debug with full visibility.

Secrets Management

Encrypted per-organization secret store. Inject API keys, tokens, and credentials securely into any node.

Extensible Node Registry

Add custom nodes by extending the registry. Write once, reuse everywhere across your organization.

Pre-built Templates

Jumpstart with multi-node workflow templates for common AI patterns — RAG pipelines, data scraping, report generation.

Deploy on PandaStack

Self-host PandaFlow on your own infra or deploy instantly on PandaStack.io with managed PostgreSQL and Redis included.

Open Source (MIT)

Fully open source under the MIT license. Fork it, extend it, contribute back. Built in public at github.com/pandastack-io/pandaflow.

Quick Start

Run PandaFlow locally in under 5 minutes.

1

Clone the repo

git clone https://github.com/pandastack-io/pandaflow.git
2

Install dependencies

npm install
3

Copy env file

cp .env.example .env.local
4

Run migrations

npm run db:migrate
5

Start dev server

npm run dev

Requires: Node.js 20+, PostgreSQL 16, Redis 7

Tech Stack

Built with modern, production-grade technologies.

FrontendNext.js 16, React 19, TypeScript, Tailwind CSS, React Flow
BackendNext.js API Routes, Drizzle ORM
DatabasePostgreSQL 16
CacheRedis 7
ExecutionFirecracker microVMs
AuthNextAuth v5 (email + Google OAuth)

Deploy on PandaStack.io — the easiest way to host PandaFlow with managed databases, automatic scaling, and Firecracker microVM execution out of the box.

Open source. Contributions welcome.

PandaFlow is MIT licensed and built in public by PandastackIO Inc. Star the repo, open issues, and help shape what comes next.