Think Dockerfile or SQL for multi-step LLM pipelines: you declare what to do, not how. The runtime handles execution across any model or provider. The Pipelex language (.plx) is an open standard (MIT license) designed to be both human-readable and LLM-friendly.

Quick links: Website • Docs • Discord • Cookbook • Demo (2min) • VS Code Extension

• pipelex — Runtime & PLX language The Python library and language specification. Build, run, and compose AI workflows with structured I/O and validation.

• pipelex-api — FastAPI server & Docker 🚀 Deploy Pipelex as a REST API. Self-host with Docker or integrate into your infrastructure.

• pipelex-mcp — Model Context Protocol server 🤖 Let Claude, Cursor, and other MCP-compatible agents run and build Pipelex workflows.

• n8n-nodes-pipelex — n8n automation node ⚡ Integrate Pipelex workflows into n8n for no-code automation and orchestration.

• pipelex-cookbook — Examples & recipes Ready-to-run pipelines: classification, extraction, analysis, generation, and more.

• pipelex-starter — Project template Bootstrap new projects with batteries included: Makefile, tests, environment setup.

LLM Access: Use Pipelex Inference (free tier available) for unified access to OpenAI, Anthropic, Google, Mistral, and more—or bring your own API keys.

pip install pipelex
# Optional: provider-specific extras
pip install "pipelex[anthropic,google,mistralai]"

from pipelex import run_pipeline

result = run_pipeline("my_workflow.plx", {"input": "data"})

# Using Docker
docker pull pipelex/pipelex-api:latest
docker run -p 8081:8081 \
  -e API_KEY=your-api-key-here \
  -e PIPELEX_INFERENCE_API_KEY=your-pipelex-key \
  pipelex/pipelex-api:latest

# Or clone and run locally
git clone https://github.com/Pipelex/pipelex-api
cd pipelex-api
docker-compose up

# Install the MCP server
pip install pipelex-mcp

# Configure in Claude Desktop or Cursor
# Agents can now run AND build Pipelex workflows

Install the Pipelex node in n8n to integrate AI workflows into your automation pipelines. Connect with 400+ apps and services.

git clone https://github.com/Pipelex/pipelex-cookbook
cd pipelex-cookbook
make install
cp .env.example .env  # Add your API keys

# Run an example
python examples/classification/sentiment.py

# Use GitHub template
# Visit: https://github.com/Pipelex/pipelex-starter
# Click "Use this template" → Create your repo

git clone <your-new-repo>
cd <your-new-repo>
make install

# Install Pipelex with CLI
pip install pipelex

# Describe what you want, get a working pipeline
pipelex build pipe "Extract product features from reviews and rate importance 1-10"

# Run the generated pipeline
pipelex run product_features.plx --input "review.txt"

💡 Pro tip: Install vscode-pipelex for syntax highlighting in VS Code, Cursor, Windsurf, and BlackboxAI.

We welcome contributions! Here's how to get involved:

• 💬 Discord — Get help, share workflows, meet the team
• 📚 Documentation — Complete guides and API reference
• 🐛 GitHub Issues — Report bugs and request features
• ✉️ security@pipelex.com — Security and privacy concerns

All repositories are MIT licensed unless otherwise specified. See individual LICENSE files for details.

"Pipelex" is a trademark of Evotis S.A.S.

Built by developers who were tired of rewriting the same AI patterns.
Now you don't have to.