Solo Ops Lab

Automation for one-person businesses: Zapier vs Make vs n8n cost math, what breaks, and when you don't need AI.

[Solo Ops Lab @SoloOpsLab]

Most automation advice is written by people selling the software.

This account does the unglamorous part: what a workflow actually costs on Zapier, Make, and n8n, where it tends to break, and when a plain rule beats an AI step.

Sources linked. No income screenshots.

[Solo Ops Lab @SoloOpsLab]

@just_tesaa Short version: Zapier charges a task per successful action step and has the most apps. Make charges per module run, and a polling trigger costs an operation every check. n8n charges per whole workflow run and you can self host it. Long flows favor n8n.

[Solo Ops Lab @SoloOpsLab]

@rodri_shema Worth adding under point 1: whoever maintains a self hosted n8n also owns the upgrades. The stable line got two patch releases today alone, 2.47.6 and 2.47.7. If nobody on the team will read release notes, that's a vote for Zapier or Make.

[Solo Ops Lab @SoloOpsLab]

@johnaxelsson_ That Zapier number usually comes in lower. Triggers never use tasks, and neither do Filter, Paths, Formatter, Delay or Looping steps. Steps that error or halt don't count either. Count only the successful action steps in the loop.

[Solo Ops Lab @SoloOpsLab]

@zapier Also on the live board: Hydrogen at medium effort is 46.73% for $0.41 a task. Half the cost again for about 4 points less. For a high volume, low stakes step like tagging or routing, that's the cheaper one to test first.

[Solo Ops Lab @SoloOpsLab]

n8n fix: an Agent on an OpenAI credential with a custom base URL, like LiteLLM in front of Bedrock, could fail every run on prompt_cache_key.

2.47.6, out today, adds a Prompt caching switch in the agent's Advanced panel. Update, then turn it off for that agent.

[Solo Ops Lab @SoloOpsLab]

@TheConnMan If Claude is managing failures, the n8n MCP server now lets it connect an error workflow in workflow settings and restore an earlier version from history (2.29). That covers a lot of failure handling without opening the editor.

[Solo Ops Lab @SoloOpsLab]

@Gstielz For the bookings branch: n8n can gate a single tool behind human approval. On the AI Agent, click + on that tool's connection and pick Add human review step. The agent still answers FAQs on its own, but asks before it writes a booking.

[Solo Ops Lab @SoloOpsLab]

@Atharv_Golait For a business that grows, n8n, mostly on billing. One execution is a whole workflow run, however many steps. Zapier charges a task for each successful action step. And n8n error workflow runs no longer count toward your quota, so there's no reason to skip alerts.

[Solo Ops Lab @SoloOpsLab]

n8n change: since 2.45, a Schedule Trigger set to every N minutes runs N apart from when you published it, not on the clock. Every 15 no longer means :00, :15, :30, :45.

If a report or sync needs clock times, switch to Custom (Cron) with */15 * * * *. Cron keeps its timing.

[Solo Ops Lab @SoloOpsLab]

@riglabs_ The $5 check in is in n8n's launch post, and you can change that amount. Two other limits from the same post: the permission check is itself a model and can make mistakes, and Nod stops when your laptop sleeps. Scheduled jobs need the machine awake, or a small always on box.

[Solo Ops Lab @SoloOpsLab]

@zapier Worth knowing this is Next Gen Zaps, early access, listed on Team and Enterprise. If you get it, stay on Suggest fixes for any Zap that emails a client or moves money. Automatically fix can publish a new version after a test, and a small code change can go out on its own.

[Solo Ops Lab @SoloOpsLab]

@zapier @AnthropicAI 51.4% is the best Anthropic score, but the live board at https://t.co/NIzuJn5eaa still has almost half failing. Fable 5.2 at max effort costs $4.64 per task. Gemini 4 Carbon at high effort is 53.88% at $2.12. For a solo shop, don't default to the most expensive model.

[Solo Ops Lab @SoloOpsLab]

n8n tip: if a webhook gets health checks or events you don't care about, each one still starts an execution, even if the flow quits at the first IF.

Webhook node, Add Option, Only Run If. Requests that fail the expression get a 200 and never create an execution.

[Solo Ops Lab @SoloOpsLab]

@gustavonicot @elonmusk @bot it does. seems if you're using a connector the bot might be blocked but otherwise my bot has its own local server it executes tasks on.

[Solo Ops Lab @SoloOpsLab]

Zendesk heads up: since Oct 27 you can't create new API tokens, and a token unused for 30 days gets deactivated. All of them stop working April 30, 2027.

If an HTTP step in Make or n8n calls Zendesk with a token, especially one that only runs monthly, move it to OAuth now.

[Solo Ops Lab @SoloOpsLab]

n8n tip: if your instance is down when a Schedule Trigger is due, that run is skipped by default.

In the node's Settings tab, set If Execution Is Missed to Run the Most Recent Missed Execution for one catch up run.

Needs a node added in 2.36+ and the durable scheduler.

[Solo Ops Lab @SoloOpsLab]

Zapier tip: export up to 12 completed billing periods of task usage as a CSV from Billing and usage, then Usage insights.

Each row shows tasks used next to your plan limit. Open beta on Free, Professional, and Team with a monthly task limit. Not on Enterprise or annual limits.

[Solo Ops Lab @SoloOpsLab]

Make tip: Gemini 3.5 Flash and 3.7 Flash are off the model lists in Generate a response, Extract structured data, and Simple text prompt.

Your running scenarios still work. Google routes those to 3.6 and 3.8. If you edit the module, pick a model again before you save.

[Solo Ops Lab @SoloOpsLab]

Make tip: each bundle a trigger returns runs the next module again. Watch a sheet, get 10 new contacts, send one email each: 11 operations.

Put a Text aggregator before the email and send one digest instead. Make's own docs example drops that run to 3 operations.

[Solo Ops Lab @SoloOpsLab]

Zapier tip: in a Classic Zap, Filter, Paths, Formatter, Delay, Looping, Digest and Storage steps never use tasks. Neither do triggers or steps that error.

One catch: starting December 1, a Next Gen Zap run that only filters or formats still costs 1 task.