←BaldanAll skillsLet's talk

Automation & AI Agents · Tools

Every tool for the job, and when to pick it.

No-code platforms, robots, browser bots, schedulers, orchestrators, chat APIs, models and agent frameworks: 71 tools in 10 layers, the 7 I build with among them.

My automation stack

PAPower AutomateNo-code & iPaaSPWPlaywrightBrowser & testingChrome extensionsBrowser & testingPythonScripts & schedulersNode.jsScripts & schedulersWhatsApp Cloud APIMessagingModel Context ProtocolAgent protocols, memory & ops

01 · The map

71 tools, ten layers.

From no-code platforms and robots to queues, models and agent frameworks: what people actually use to automate, and what each is for. The 7 marked with a dot are the ones I build with. Click any tool.

No-code & iPaaS

connect apps with boxes and arrows

RPA

robots that use screens, for systems without APIs

Browser & testing

code that drives a real browser

Scripts & schedulers

languages and timers for jobs you write

Orchestration & queues

long, multi-step and background work, reliably

CI/CD & infrastructure

shipping code and building servers automatically

Messaging

where automations talk to people

AI models & gateways

the models that read, write and decide

Agent frameworks

code for loops, tools, memory and teams of agents

Agent protocols, memory & ops

connecting agents to tools and data, and watching them

02 · Same task

One job, written in many tools.

The best way to see the difference between tools is the same job done in each. Pick a job, then flip between tools: the code, the steps and the trade-offs change; the result doesn't.

report.py · cron: 0 7 * * *

14 lines
import datetime as dt, os, requests

day = (dt.date.today() - dt.timedelta(days=1)).isoformat()
orders = requests.get(f"{os.environ['API']}/orders",
                      params={"date": day}, timeout=30).json()

totals = {}
for o in orders:
    totals[o["store"]] = totals.get(o["store"], 0) + o["total"]

lines = [f"{s}: Rp {t:,.0f}" for s, t in sorted(totals.items())]
requests.post(os.environ["CHAT_WEBHOOK"],
              json={"text": f"Orders {day}\n" + "\n".join(lines)},
              timeout=30)

Run by cron on any Linux server: 0 7 * * * python3 /opt/jobs/report.py

04 · Head-to-head

Two tools, side by side.

Pick any two, or start from a classic pairing. Tools from different layers can be compared too: often the real question is no-code or code, robot or API, workflow or agent.

Easy to start
Handles complex logic
Connectors & community
Self-host & control
Grows with you

Zapier

No-code & iPaaS · No code

The best-known no-code automation service: a trigger in one app, actions in others.

+ Thousands of app connectors and the gentlest learning curve.

− Priced per task; branching and loops get awkward and expensive.

Pick it for small teams connecting SaaS apps without a developer.

n8n

No-code & iPaaS · Low code + JavaScript

Open-source workflow automation you can host yourself, with code nodes and AI agent nodes.

+ Self-hostable, so data stays with you; code when you need it; strong AI building blocks.

− Self-hosting means you run, update and back it up; fewer connectors than Zapier.

Pick it for teams that want control, privacy or complex logic, and AI agents inside workflows.

Ratings are my rule of thumb for typical use, not benchmarks.

05 · Pick a stack

Where would you start?

Seven common situations and a sensible starting stack for each, with the reason behind every pick and what to use instead. A starting point, not a rule.

The situation

Orders, invoices and reminders across a few online apps, run by the owner or an admin.

No-code & iPaaS

Make

visual workflows with branches, without code

or: Zapier

Scripts & schedulers

Google Apps Script

the order sheet already lives in Google Sheets

or: Excel macros & Office Scripts

Messaging

WhatsApp Cloud APII use it

customers and staff already answer WhatsApp

or: Telegram Bot API

AI models & gateways

OpenRouter

one key to try AI steps (sorting messages, drafting replies)

or: OpenAI API