n8n's last few releases are less about new integrations and more about AI agents you can watch and price. Here is what is verifiable, and how it compares with Zapier and Make.
According to n8n's official release notes, version 2.43 (dated October 6, 2026) includes an entry titled "Release Browser Use to all users." The notes say Browser Use in the n8n Assistant is now available to all users, controlled by an instance admin setting that is on by default, and that it is unavailable on phones and tablets. The same release adds a deployment dropdown for the Azure OpenAI Chat Model node, OAuth credentials for the Supabase node, and keyboard controls for the Assistant panel.
The two previous releases matter for cost. Version 2.42 (September 29) added prompt caching to the Anthropic node's Message operation, and the notes also list a write lock for the agent builder. Version 2.40 added a session details panel in the agent preview chat that shows token usage, cost and duration. Prompt caching generally means repeated input is billed at a reduced rate, but it does not cap spending.
One caution: a third-party release tracker describes 2.43 as also adding a budget guardrail for agent runs. We could not find that in n8n's official release notes, so we are not treating it as confirmed. Check the release notes yourself before relying on any spending limit.
The risk with AI agents is not just whether they work. It is whether you can see what they cost. Visibility into tokens, cost and duration per session is a practical step toward answering that, particularly when you pay a model provider directly with your own API key. Whether that visibility translates into real savings depends on how you build your workflows.
| Tool | Billing model | Best fit |
|---|---|---|
| n8n | Free self-hosted community edition; cloud plans reportedly start around $20/mo (verify on n8n's pricing page) | Technical users who want control and are fine managing hosting |
| Make | Credit-based; paid plans reportedly start around $9/mo billed annually | Visual, multi-step workflows without code |
| Zapier | Task-based | Beginners who want the widest app support and simplest setup |
Pricing figures are from public sources and may be out of date. See our Make vs Zapier breakdown and the Zapier AI task pricing explainer for the managed-platform side.
The community edition costs nothing in license fees, but you carry hosting, updates, security and backups. n8n describes its licensing as source-available (a "fair-code" model), so read the terms if you plan to use it commercially.
Letting an AI assistant operate a browser can save time on repetitive web tasks, but it can also click the wrong thing. Start with low-stakes tasks, avoid giving it access to payment or admin accounts, and review what it does before expanding scope.
n8n has been shipping weekly releases, with several recent ones focused on the AI assistant and agent builder. That is good for features and tiring for anyone who wants a stable setup. Pin versions and read release notes before upgrading.
If you have a developer or a technically confident operator, and you are running enough AI-heavy workflows that per-step platform fees add up, n8n is worth a trial on a small project. If you do not, the managed platforms remain the lower-effort option, and the cost visibility gap is worth watching rather than switching over.
n8n's recent releases show a steady push toward AI agents that are easier to observe and a bit cheaper to run. Browser Use reaching everyone is the headline, but the cost-tracking changes are the more useful detail for budget-minded teams. Confirm any spending controls in the official docs before depending on them.
Note that StackPilot's comparisons are based on research of public information and vendor documentation, not hands-on testing. Pricing and release details may change; confirm with each vendor.