Orkeon
Orkeon Studio

The team composed in Studio is the one the server runs.

Studio produces a team folder in YAML, which the CLI runs as is.

Getting started on Windows

  1. Install the orkeon CLI

    Install the CLI

  2. Download Orkeon Studio

    Download Studio (.msi) (new tab)

    The installer is unsigned: Windows warns at first launch. First compare its checksum with the one in the SHA256SUMS file published next to it.

  3. Tell Studio where the CLI is

    After the Orkeon installer, set the environment variable ORKEON_CLI_DIR to %LOCALAPPDATA%\Programs\Orkeon\libexec\orkeon: without it, Studio does not find the CLI.

Kama in a blacksmith’s apron, goggles on the forehead, hammer in hand in front of an anvil, surrounded by molecules
Key points

Creating a team without code or a terminal, in four steps

Orkeon Studio is a desktop application for Windows that drives the orkeon CLI, which must be installed first.

In the application, the steps are called Describe → Compose → Try → Adopt.

  1. Describe the need

    You say what the team must do.

  2. Compose the team

    Studio proposes the agents, their tasks and the folders they are allowed to touch.

  3. Try it on a sample

    The trial is optional, and Studio shows its cost.

  4. Adopt, on your decision

    The adopted team becomes a folder on your disk.

Two terminal applications, orkeon-studio-config and orkeon-studio-run, run on Linux, macOS and Windows; only the Linux x64 archive is published (new tab).

In detail

How Studio works, and its limits

  1. The CLI runs as a child process of Studio

    Studio starts the CLI, shows its output and produces the files it runs.

  2. Studio is neither a scheduler nor a NuGet package

    A team to start every morning is handed to the operating system’s scheduler, with your consent.

  3. The team folder is copied as is to a server

    The folders opened to the agents must be authorized again there; the accounts and the model profile are configured there.

  4. Web tools are declared like the others

    A team that receives them reaches the whole public web until the domain list is declared.

  5. Studio shows the cost of each trial and each run

    There is no cost cap by default (Security).

  6. An agent can be wrong

    Human review remains necessary where the stakes call for it.

  7. What an agent reads is sent to the model

    With a cloud provider, this data leaves the company; with a local model, it stays there.

Further reading

Studio’s repository and documentation