The Pipelex plugin for Claude Code and Codex: build and run AI methods from your agent, with skills that write them, a hook that checks every edit, and the Pipelex tools.
Pipelex runs your AI methods — write a method once, then run it from your agent via MCP, turn it into a webapp, or use it 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 Code
claude plugin marketplace add Pipelex/pipelex-plugins
claude plugin install pipelex@pipelex-pluginsClaude 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; nothing else to install.
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
codex plugin marketplace add Pipelex/pipelex-plugins
export PIPELEX_API_KEY=plx_sk_... # create one in your console at app.pipelex.comRestart Codex, run /plugins to install pipelex, and trust the plugin hook on first run. Requires Codex 0.141 or later.
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.pdfand 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
This repository is the Pipelex plugin — what it holds, the other agents it installs in and how to work on it are below.
Your agent picks the skill your request calls for, and you can also name one yourself.
/pipelex-designdesigns a method contract-first and writes it as.mthdsfiles, directly when the method is simple and step by step when it is not./pipelex-editmakes an edit that keeps a method's contract, such as a prompt, a model or a rename, and proves it with a validation before and after./pipelex-organizeregroups a method's files into a layout you can browse, without changing what the method does./pipelex-explainexplains a method in plain language, its contract, its flow and every pipe, and writes nothing./pipelex-inputsprepares a method's inputs from your files, from synthetic data or from a template, ready to run./pipelex-synthetic-inputsmakes the test files a method needs when you have none: PDFs, images, Word and Excel files rendered from code, and photographs from an image model./pipelex-runruns a method on the hosted Pipelex API, prints its run id at once, and follows a run to its status, its results and its files./pipelex-catalogsaves a method to your Pipelex account, lists what is saved there, and pulls a saved method back to disk./pipelex-labframes a use case into candidate methods, writes each test case's answer before the first run, then runs, scores, logs and fixes the method round after round within a budget you agree./pipelex-integratewires a method into your TypeScript or Python code, with generated types and one typed call through@pipelex/sdkorpipelex-sdk./pipelex-scaffoldstarts a new project around a method: a web app from the method-app template, or any other project from its language's own initializer.
The skills share a reference for MTHDS, the language a method is written in. Each skill in detail, and what it needs.
After every edit to a .mthds file, the hook lints it and formats it in place on your machine, offline, then validates the whole method on the hosted Pipelex API when a key is set. A failed check returns to your agent with the line to fix. How the hook works.
The tools the skills call to validate a method, prepare its inputs, run it, generate typed code and reach your saved methods start with your agent, with nothing else to install. The tools, one by one.
The plugin also installs in Mistral Vibe 2.21.0 or later, from a clone of this repository: Vibe loads the skills from a path in its config, the hook from its hooks.toml, and the Pipelex tools from a config entry that carries your key. The Mistral Vibe procedure.
Claude Code in the Claude desktop app does not inherit your shell environment, so an exported PIPELEX_API_KEY never reaches it. Set the key in the plugin configuration dialog, which opens when you enable the plugin.
Which Pipelex product each app takes, and every agent's install in detail, are in Install the plugin, agent by agent.
- API key. In Claude Code, enter it in the plugin configuration dialog, which stores it in your OS keychain;
PIPELEX_API_KEYin your environment is the fallback, and a value set in the dialog wins. Codex readsPIPELEX_API_KEYfrom your shell. Create a key in your console at app.pipelex.com. - Base URL. Optional, for a deployment other than the hosted API: the dialog's base URL field, or
PIPELEX_BASE_URL. It defaults tohttps://api.pipelex.com. - Without a key, the hook still lints and formats every edit and skips only the validation, and the skills that need the Pipelex tools stop and say how to set one. The hook never blocks an edit because Node.js is missing or the API is unreachable.
- Privacy. With a key set, each validation sends the
.mthdsfiles around the edited one to the Pipelex API. Leave the key unset, in the dialog and in your environment, to keep the hook's checks on your machine.
- Install the plugin, agent by agent: Claude Code, Codex and Mistral Vibe in detail, and which Pipelex product each app takes.
- The skills and the Pipelex tools: what each skill does and needs, and the tools it calls.
- The validation hook: the checks it runs, on each agent, and what it does when something is missing.
- Develop the plugin: the build, the checks, and running your changes in your own agent.
- The multi-target build, continuous integration and the decisions behind the plugin.
- The Pipelex documentation and the MTHDS standard.
The plugin is rendered from the Jinja2 templates in templates/ into one output per agent. Edit the templates, never the generated pipelex/, pipelex-codex/ or pipelex-vibe/ trees.
make build # render every target
make agent-check # fix imports, format, lint, then the checks CI runs
make agent-test # the unit tests, quiet unless one failsTo run your changes in Claude Code or Codex before they are released, see the dogfood loop.
Apache 2.0. See LICENSE.