TL;DR
- Managed Agents bills at $0.08 per session-hour. N8N self-hosted is $20/month. Different agent cost models entirely.
- Flat monthly cost suits predictable, always-on workflows. Usage-based cost suits bursty or uncertain session volume.
- Neither is wrong. Picking the wrong one for your load profile is.
Managed Agents costs $0.08 per session-hour. N8N self-hosted costs $20/month. They both run automation logic, but the agent cost structure and the operational tradeoffs are completely different.
What does $0.08 per session-hour actually mean for agent cost?
Managed Agents charges by active session time, so your agent cost scales directly with usage.
If your agent sits idle, you pay nothing. If it runs hard across many concurrent sessions, the bill climbs. That's the deal. For businesses with unpredictable or spiky volume, this looks attractive early. You're not paying for capacity you haven't used yet. But if your agent runs long sessions frequently, the per-hour rate compounds. It's not a trap. It's just math you need to model before you commit.
A practical way to think about it: map out your busiest expected day. Estimate how many sessions run, and for how long. Multiply that by $0.08 and then scale it to a month. If that number sits comfortably inside your budget even at peak volume, the model works for you. If peak volume makes the number uncomfortable, a flat-rate alternative deserves a closer look. The modelling takes less than ten minutes and it removes most of the uncertainty before you build anything.
Why is N8N self-hosted at $20/month a different product entirely?
N8N self-hosted at $20/month gives you a flat infrastructure cost regardless of how many workflows you run.
You're paying for a server, not for execution time. Run one workflow or a hundred. The agent cost stays predictable. That predictability is what makes it useful for production builds with steady, known volume. The tradeoff is you own the ops. Updates, uptime, backups. That's on you or your builder. It's not a managed service. For an indie builder running client systems, that's usually fine. For an SMB owner who doesn't want to think about servers, it's a different conversation. We've built 18 N8N workflows for non-technical clients and the self-hosted model held up well, but only because someone was watching it.
The $20/month figure also assumes you're running on modest infrastructure. If your workflow volume grows and you need to scale the underlying server, that base cost moves. It's still predictable in structure, but not permanently fixed. Factor that into your planning if you're expecting significant growth in the first year.
How do you pick the right agent cost model for your build?
The right choice depends on your session volume, your ops capacity, and how predictable your workload is.
Here's a plain breakdown of the tradeoffs:
- Variable volume, low ops appetite: Managed Agents. You pay as sessions happen and someone else handles the infrastructure.
- Steady volume, builder on deck: N8N self-hosted. Flat agent cost, full control, but you carry the maintenance burden.
- Uncertain volume, tight budget: Model it before you pick. At $0.08 per session-hour, high-frequency long sessions add up fast.
- Non-technical client, no builder retained: Neither option is plug-and-play. Scope the support model first.
- Regulated industry (finance, insurance): Check your data residency before choosing a managed service. ACMA compliance obligations don't disappear because your agent runs in the cloud.
Where does agent cost fit inside a real production stack?
Agent cost is just one line in a larger infrastructure bill, and it's rarely the biggest one.
For a typical voice AI build on Retell AI and GHL, the orchestration layer is one cost. Voice minutes are another. CRM seats, SMS, and any outbound calling compliance tooling all sit on top. N8N self-hosted at $20/month is close to invisible in that stack. Managed Agents at $0.08 per session-hour could be significant or negligible depending on how the agent is used. The point isn't which number is smaller. It's which cost structure fits your usage pattern. We wrote about agent pricing logic in more detail when we broke down why we picked our own pricing model. Same thinking applies here.
Does switching between the two agent cost models matter mid-build?
Switching mid-build is possible but painful, and the agent cost difference alone rarely justifies the rework.
The architectures aren't interchangeable. N8N self-hosted uses a node-based workflow model. Managed Agents has its own session and state management layer. Migrating means rebuilding logic, not just changing a config. If you're early enough that nothing's in production, the choice is worth thinking through carefully. If you've already shipped and it's working, optimise the cost model around your actual usage before you consider rearchitecting. Most operators don't need to switch. They need to understand what they're running.
Key Takeaways
- Managed Agents at $0.08 per session-hour and N8N self-hosted at $20/month are different products with different agent cost structures, not direct competitors.
- Flat monthly cost suits predictable workloads where someone manages the server. Usage-based agent cost suits variable or bursty session volume.
- Model your session frequency and duration before committing. The maths is simple and skipping it is expensive.
If you're building voice AI or workflow automation for an Aussie service business and you're not sure which agent cost model fits your stack, DM AUDIT and I'll send you five questions that'll sort it out in under ten minutes.
Originally published at theautomate.io.
Top comments (0)