Quickstart: Run your first workflow in AGNT5 Cloud

Create a workflow locally, connect it to an AGNT5 Cloud dev environment, invoke it, and inspect the execution trace.

In this quickstart you’ll create a workflow locally, connect it to an AGNT5 Cloud dev environment, invoke it, and inspect the execution trace. The workflow itself summarizes the top Hacker News stories — the API is public so there’s no Hacker News token to chase; you just bring your own OpenAI key.

You’ll need: Python 3.12+ Node.js 20+ (with pnpm, npm, or yarn) , an OpenAI API key, and ~3 minutes. We install the agnt5 CLI below.

Using a coding agent?

Paste the prompt below into Claude Code, Cursor, Windsurf, Cline, or Copilot. The agent walks every step on this page for you. If the AGNT5 MCP server is connected, it’ll also inspect the run and trace once the workflow finishes — no clicking through Studio.

Installs the CLI, scaffolds the project, starts agnt5 dev, and walks you through the first run.

Running the manual commands yourself? Skip this — the same five steps are below.

Install the CLI

Install agnt5

curl -LsSf https://agnt5.com/cli.sh | bash

brew install agnt5/tap/agnt5

The installer writes agnt5 to ~/.agnt5/bin and adds it to your PATH. Open a new terminal (or source your shell’s rc file) so agnt5 resolves.

Authenticate

agnt5 auth login

Opens a browser window to sign you into AGNT5. Once it returns, future agnt5 commands run against your account. For verification, troubleshooting, and API-key auth, see the full Install guide.

Run it

Create the project

agnt5 create --template python/quickstart my-agnt5-quickstart
cd my-agnt5-quickstart

agnt5 create --template typescript/quickstart my-agnt5-quickstart
cd my-agnt5-quickstart

agnt5 create downloads the template, registers the project with the Control Plane, and writes the scaffolded files into my-agnt5-quickstart/.

Start the dev server

agnt5 dev

agnt5 dev starts a local worker, registers your components with the runtime, and prints a Studio URL:

Registered components: digest, fetch_top_ids, fetch_story, summarize, assemble_digest
Worker connected
Studio: https://app.agnt5.com/anon/<session-id>
Watching project files

Registered components: digest, fetchTopIds, fetchStory, summarize, assembleDigest
Worker connected
Studio: https://app.agnt5.com/anon/<session-id>
Watching project files

Open Studio

Open the Studio URL from the terminal in your browser. The components your worker just registered show up live — the digest workflow is at the top of the list.

Run the digest workflow

In Studio:

Pick the digest workflow.
Set the input to {"limit": 5}.
Click Run.

The trace renders live as each step lands. Click any step to inspect its input, output, and (for model calls) the prompt, response, and cost.

Notes

Default model is openai/gpt-5-mini. Change it on the model="..." line in functions.py functions.ts .
Side effects go through ctx.task(...). A bare await fetch_story(id) await fetchStory(id) would run every replay and break resume.
You can also invoke the workflow from the CLI instead of Studio: agnt5 run digest --input '{"limit": 5}'.

What’s next

This gets you through the first part of the AGNT5 loop: Build → Ship → Run → Observe. For production-ready behavior — promoting to managed environments, invoking from your app, capturing failures into evals — work through The Loop.