How to choose an n8n agency

Start with the kind of agency you need

Agencies that all say they do n8n differ widely in what they deliver around it. Pick the type before you compare names.

Questions to ask on the first call

1. Who will build our workflows, and can we speak to that person before signing?
2. Can you show a production workflow of similar size, with client details removed?
3. How do you handle credentials and access to our systems?
4. Which n8n edition and hosting model does your proposal assume, and why?
5. What happens when a workflow fails at night or over a weekend?
6. What do we receive at handover, and can our team change the workflows afterwards?
7. How do you price changes after launch?

Self-hosted and cloud competence

n8n runs either as n8n Cloud or on your own servers. A good agency explains the trade-off for your case and does not push one option for every client. Cloud removes server work. Self-hosting gives control over data location and network access.

If you self-host, ask how the agency handles backups, updates, secrets and scaling. For higher volumes, n8n supports queue mode, where a main instance hands executions to workers through Redis and a shared database. An agency that proposes self-hosting should be able to say when you would need that and who maintains it.

Error handling and monitoring

Every automation fails at some point. What matters is whether someone finds out and whether data is lost. n8n lets you assign an error workflow that runs when an execution fails, using the Error Trigger node described in n8n's error handling docs.

Ask to see how the agency uses it. Good answers include alerts to a named channel, retries for flaky APIs, a record of failed items that can be replayed, and a clear owner for each alert. A vague answer here predicts trouble more reliably than anything else on the call.

Documentation and handover

You should be able to leave the agency without rebuilding. Put these deliverables in the contract.

• Workflows exported and stored in a repository or location you own
• Accounts, servers and credentials registered in your company's name
• A short document per workflow: purpose, trigger, systems touched, known limits
• Named nodes and sticky notes inside the workflows, so the canvas explains itself
• A recorded walkthrough or training session for the person who will maintain them

Red flags

• Hosting on the agency's own account. If the instance and credentials belong to the agency, leaving becomes expensive.
• No questions about your process. A quote that arrives before anyone has asked about exceptions and volumes is a guess.
• Only demo workflows. Template galleries and video tutorials are not evidence of production work.
• No plan for failures. If monitoring is described as an optional extra, expect silent errors.
• Implied endorsement. A directory listing is a listing. Agencies that present it as a guarantee from n8n are overselling.