Quickstart

This walkthrough creates a tiny agent, deploys version 1, and waits for its first result.

Before you begin

You need:

Node.js 20 or newer
access to a Magmell account
the Magmell CLI built from the current repository package

Until the CLI is published to a package registry, build and link it locally:

cd clients/magmell
npm install
npm run build
npm link

Verify the executable:

magmell --help

magmell login \
  --url https://api.magmell.siloga.cloud \
  --email you@example.com

The CLI prompts for your password without echoing it and stores the resulting user session in your local Magmell configuration.

2. Create a handler

mkdir hello-agent

import siloga_agent


def handle(input: dict) -> dict:
    name = input.get("name", "world")
    siloga_agent.log("creating greeting", name=name)
    return {"message": f"Hello, {name}!"}

Magmell supplies the siloga_agent runtime in the execution environment. You do not add it to the deployment requirements.

3. Deploy it

magmell deploy ./hello-agent \
  --name hello-agent \
  --version 1 \
  --wait

--wait follows the asynchronous build and exits only when the deployment is ready or its build fails. A successful command prints the deployment ID and ready.

4. Run it

magmell run hello-agent \
  --input '{"name":"Ada"}' \
  --wait

The CLI resolves hello-agent to its latest active version, waits for the run, and prints:

{"message":"Hello, Ada!"}

5. Inspect its timeline

List recent runs to find its number:

magmell runs list --deployment hello-agent

Then inspect the gateway phases and your handler log:

magmell runs events hello-agent#1

You now have an immutable deployment that can serve additional runs without another build.