For developers (technical)

Your APIs in your coding tools, and jobs on your own machines

Use the APIs you connected from your coding assistant or any MCP client, and let agents run jobs on machines you control.

Who can use the weather gateway: anyone with a key, with its default key listed and a button for a new one.
Recorded in the staging demo workspace, using public APIs.

How it fits together

The same actions, inside the tools you code with

Who it is for

Developers and platform teams. This is the technical path: the other examples do not need any of it.

The weather service the operations brief uses is also published on a gateway in the same workspace. The gateway has its own address, and anyone with one of its keys can add it to an MCP client and call the same forecast action from their editor.

For work that has to happen on a particular computer, you start a runner on that machine. You decide who may use it, and agents send it work by label, like os=linux.

Set up in the staging demo workspace. Your gateway, keys and machines will differ.

What the agent does

  • Turns an OpenAPI, GraphQL, SOAP or gRPC description, or an npm package, into tools
  • Serves them to coding assistants over MCP, or as Agent Skills or UTCP
  • Runs jobs on any machine you control, matched by labels like os=linux
  • Answers on an OpenAI-compatible API, and from a command line tool

What you need

  • An API connected in almyty
  • A coding assistant or editor that speaks MCP
  • For a runner: a machine with Node.js where you can install a package

Set up your own

Three steps to a version of this.

The docs walk through an agent of this kind step by step. It is a guide to the pattern, not a copy of this exact agent.

Read the guide
  1. Publish your API's tools

    Create an MCP gateway for the API and make a key for it.

  2. Add it to your coding tool

    Paste the address and key into your MCP client's settings. Every call shows up in Analytics.

  3. Start a runner

    On the machine, install the almyty command line tool, log in and start the runner. Then let an agent send work there by label.

Keep exploring

Other recorded examples.

Start with a job like this one.

Connect the service, describe the result you want and check the first run yourself.