pipelex

Quick start

Pipelex lets you build AI methods with your coding agent and run them anywhere — from your agent or your chatbot via MCP, as a webapp, or via API in any software.

1. Sign up at app.pipelex.com.

2. Install the Pipelex plugin in your coding agent. The plugin is how you build methods: it gives your agent the skills that write and run them, a hook that checks every edit, and the Pipelex tools.

claude plugin marketplace add Pipelex/pipelex-plugins
claude plugin install pipelex@pipelex-plugins

Claude Code asks for an API key when you enable the plugin, and stores it in your OS keychain — create one in your console at app.pipelex.com. The skills, the hook that checks every edit and the Pipelex tools load with it. The plugin's hook and the Pipelex tools run on Node.js, so you need Node.js on your PATH.

Claude Code also loads what you have added to your Claude account, so if the Pipelex MCP is there, turn it off in Claude Code with /mcp: an agent with the plugin never takes both, since they register the same tool names.

codex plugin marketplace add Pipelex/pipelex-plugins
export PIPELEX_API_KEY=plx_sk_...     # create one in your console at app.pipelex.com

Restart Codex, run /plugins to install pipelex, and trust the plugin hook on first run. Requires Codex 0.141 or later. The plugin's hook and the Pipelex tools run on Node.js, so you need Node.js on your PATH.

3. Ask your agent for the method you want.

Design a method that reads an invoice PDF and returns the supplier, the total and the line items. Then run it on ~/Downloads/invoice.pdf and save it to my Pipelex account.

/pipelex-design writes the method, the hook checks it on every edit, /pipelex-run starts it and prints a run id you can come back to, and /pipelex-catalog saves it to your account, where your chatbot can run it too.

Run your methods from your chatbot. The Pipelex MCP is a connector for your chatbot (ChatGPT, Claude): it gives it access to the Pipelex service, so it can list the methods saved in your account and run them right in the conversation. Your methods become your chatbot's tools. To build a method, use the Pipelex plugin in a coding agent such as Claude Code or Codex, as in steps 2 and 3 above.

Add the Pipelex MCP in your chatbot's settings by the address below — in Claude, that is Add custom connector — then sign in with your Pipelex account when asked. Nothing to install and no key: the Pipelex MCP runs on your signed-in session.

https://mcp.pipelex.com/mcp

Then ask your chatbot:

What methods do I have?

Run the invoice method on https://example.com/invoice.pdf

You get a run id straight away, and you can ask for its status, its results or the files it produced at any time.

Give the file as a URL the Pipelex MCP can reach. In ChatGPT you can attach it to the conversation instead and ask for a run on it; Claude has no way yet to hand the Pipelex MCP a file you attached.

The other two ways. Turn the method into a webapp with the method-app template, or use it via API in any software through POST /v1/start — in TypeScript with @pipelex/sdk, in Python with pipelex-sdk, or with any HTTP client.

Next: what Pipelex is · documentation · your console · Discord

Prefer to run it yourself? This repository is the Pipelex runtime — its own install and configuration are below.

Run it yourself

This repository is the Pipelex runtime: the Python package that reads a .mthds file and runs it. Install it and everything happens on your own machine, against the model providers you choose.

Install

uv tool install pipelex
pipelex init
pipelex doctor

pipelex init writes your ~/.pipelex configuration and offers to install the editor extension; pipelex doctor reports what is configured and what is missing.

Some providers and features need an extra:

• anthropic: Anthropic/Claude support for text generation
• google: Google models (Vertex) support for text generation
• google-genai: Google Gemini API support for text and image generation
• mistralai: Mistral AI support for text generation and OCR
• bedrock: Amazon Bedrock support for text generation
• fal: Image generation through fal
• linkup: Web search with Linkup
• docling: OCR with Docling

Name the ones you need when you install, or take them all:

uv tool install "pipelex[anthropic,google,google-genai,mistralai,bedrock,fal,linkup,docling]"

Configure AI Access

• Bring Your Own Keys — Use existing API keys from OpenAI, Anthropic, Google, Mistral, etc. See Configure AI Providers.
• Local AI — Ollama, vLLM, LM Studio, or llama.cpp — no API keys required. See Configure AI Providers.

Run a method

Save the method shown under What a method looks like, further down, as summarize.mthds, and its inputs as inputs.json:

{
  "article": "Paste the text of an article here.",
  "audience": "busy executives"
}

The method names no model, so it runs on the deck's default-general alias, which pipelex init points at an OpenAI model. With a provider other than OpenAI or Azure OpenAI, point that alias at one of your provider's models first, by adding it to ~/.pipelex/inference/deck/x_custom_llm_deck.toml — the models each provider serves are listed under ~/.pipelex/inference/backends/:

[llm.aliases]
default-general = "claude-5-sonnet"     # an Anthropic model, for example

Then run it:

pipelex run bundle summarize.mthds --inputs inputs.json

The result is written under results/. For a method with several steps, typed concepts and a batch, run from the CLI and from Python, read CV batch screening, step by step.

Editor extension

.mthds syntax highlighting and flowchart visualization: the VS Code Marketplace, or the Open VSX Registry for Cursor, Windsurf and other VS Code forks. pipelex init offers to install it when it detects your IDE.

See Pipelex in action

Claude Code builds your AI method

What a method looks like

A method is a reusable, typed AI procedure, written in MTHDS, an open standard, and saved as a .mthds file. Each step is explicit, each output is structured, and every run is repeatable.

domain    = "articles"
main_pipe = "summarize_article"

[pipe.summarize_article]
type        = "PipeLLM"
description = "Summarize an article for a given audience"
inputs      = { article = "Text", audience = "Text" }
output      = "Text"
prompt      = "Summarize $article in three bullet points for $audience."

From here, Pipelex handles model routing across providers, structured output parsing, and pipeline orchestration.

Why methods?

Run anywhere

The same .mthds file runs from mu