Our read
Shortlist by governance need first. n8n and Zapier publish prices and, in different ways, their error handling, so you can test them this week. Workato and Tray.io publish more on regions, audit logs and environments for controlled deployments, but Tray.io's prices and both vendors' retry behaviour are not shown on the pages read. Pick one finalist from each pair, then compare them on one failing workflow rather than on connector counts.
Published hosting, error-handling and billing details for the four platforms; entries marked not shown were not on the pages read.
| Criterion | n8n | Zapier | Workato | Tray.io |
|---|---|---|---|---|
| Hosting and governance | n8n EU cloud on Starter/Pro; Business is self-hosted only SSO tier: security page and Business card differ | Zapier US hosting; SAML SSO from Team Other regions not shown on pages read | Workato 8 regions, but Pro is hosted in US Extra locations for Enterprise | Tray.io US, EU, APAC; regions on Enterprise Add-on for Team |
n8nStarter and Pro are hosted by n8n on Azure, data in the EU. Business is self-hosted only, so infrastructure and encryption at rest are the customer's. Security page puts SSO, SAML, LDAP in Enterprise; the Business card lists SSO, LDAP, environments, Git. SOC 3 public.[1] ZapierHosted on AWS in the United States; SOC 2 Type II and SOC 3. Allowed apps, action restrictions, approvals, audit logs and SCIM are described; the pricing page lists SAML SSO on Team.[2] [3] WorkatoEight regional data centers and a long certification list; Pro data is hosted in the US, extra locations for Enterprise subject to availability. Audit logs keep 90 days, configurable.[4] [5] Tray.ioRegional hosting in the US, EU and APAC; SOC 2 Type 2 and HIPAA. Dev, staging and prod separated by approval gates; SSO, HIPAA and regional hosting included in Enterprise.[6] [7] | ||||
| Error handling and traceability | n8n Retries and error workflows on all plans Log streaming: Enterprise only | Zapier Autoreplay, but not guaranteed Error notifications on paid plans | Workato Audit logs; retry not shown Check the documentation | Tray.io Run-level logs; streaming Enterprise only; retry not shown Team: streaming add-on. Pro/Team logs: 7 days |
n8nAutomatic retries, error workflows and execution logging on every plan; saved-execution log retention 7 days on Starter, 30 on Pro, unlimited on Enterprise.[8] ZapierPaid plans include custom error notifications and Zap version control. The security page says autoreplay and flood protection recover many failures, but replay is limited and not guaranteed.[3] [2] WorkatoAudit logs capture executions, configuration changes and user actions, with extended retention via an external SIEM. Automatic retry is not described on the pages read.[4] Tray.ioLogs every workflow run, agent action and MCP tool call. Log streaming to Splunk, Datadog or a SIEM is Enterprise, or an add-on for Team; Pro and Team keep logs 7 days (Team expandable). Automatic retry is not described.[6] [7] | ||||
| Billing unit | n8n Per complete run; 2.5k to 40k monthly From $20/month, billed annually | Zapier Per step and connector call; shared pool Rates vary by AI tier and runtime | Workato Unified credit; Pro $100 for 3.5k FAQ cites 2,500 to 50,000 monthly | Tray.io Tasks plus HCU; quote only No per-user or connection fees |
n8nOne execution is one complete workflow run regardless of steps. Starter 2.5k, Pro 10k, Business 40k (self-hosted only) per month; overage capped on Starter and Pro, custom on Enterprise.[8] ZapierEvery Zap step and external connector call uses tasks; one pool covers Zaps, AI, MCP and SDK. Free has 100 tasks; Professional from $19.99, Team from $69 per month.[3] WorkatoOne credit measures tasks, API calls, rows processed and agent actions. Free: 50K one-time credits; Pro: $100 a month with 3.5k credits; Enterprise on quote.[5] Tray.ioUsage is billed in tasks, with no per-user, per-connection or per-runtime fees; HCU appears as a further unit in its FAQ. Pro, Team and Enterprise are all by quote.[7] | ||||
Sources
What this comparison covers, and what it leaves out
The shared need is simple: connect business applications and keep the workflows healthy after launch, with access control, recovery when a step fails, and usage costs you can forecast. The four tools sit in one category, but they describe their reach differently. n8n cites 400+ integrations, Zapier 9,000+ apps, and the Workato and Tray.ai home pages cite 17,000+ and 12,000+ apps, databases and APIs. Each vendor counts in its own way, so connector totals are not used here to rank anything. Check the specific apps on your list instead.
The analysis therefore stays on three criteria each vendor documents: hosting and governance, error handling and billing units. Everything below comes from the vendors' own pricing, security and customer pages, so it shows what they publish, not how the products behave under your load.
Hosting and governance depend on the plan, not only the product
n8n's hosted plans, Starter and Pro, run on Azure with data stored in the European Union. The Business plan is labelled self-hosted only on the pricing page, so there the infrastructure and the encryption of data at rest are yours; the same applies to any Enterprise instance you choose to self-host. Role-based access comes with paid plans. On SSO and LDAP the pages differ: the security page places SSO, SAML and LDAP in Enterprise, while the pricing page lists SSO and LDAP, multiple environments and Git version control on the Business card. Its feature table has only Starter, Pro and Enterprise columns and shows these as not included for Starter and Pro. Get the tier in writing before you plan around SSO. Audit logging and log streaming to services such as Datadog are Enterprise only on the pricing page. n8n says its SOC 2 report goes to Enterprise customers and a SOC 3 report is public.
Zapier hosts on AWS in the United States and holds SOC 2 Type II and SOC 3. Its security page lists allowed-app lists, action restrictions, approvals before changes, audit logs, log streaming, SAML SSO and SCIM. The pricing page ties SAML SSO to Team and advanced admin permissions and observability to Enterprise, so ask which of those controls your tier really includes.
Workato lists eight regional data centers and a long certification list, but its pricing FAQ says Pro data is hosted in the United States and extra locations are for Enterprise, subject to availability. Audit logs have configurable 90-day retention. Tray.ai offers US, EU and APAC hosting and keeps dev, staging and production apart with approval gates; regional hosting, SSO and HIPAA come with Enterprise and are add-ons for Team. In both cases the governance depth you read about is an Enterprise conversation.
Error handling: n8n and Zapier publish more than the other two
n8n lists automatic retries, error workflows and execution logging on every plan. Limits differ by plan: saved-execution log retention is 7 days on Starter, 30 on Pro and unlimited on Enterprise, and maximum execution duration is 5 minutes on Starter versus 40 on higher plans. Those numbers matter for long-running syncs.
Zapier's paid plans include custom error notifications and Zap version control. Its security page says autoreplay and flood protection recover many failures, but replay is limited and not guaranteed. Do not assume a replay completes a half-finished run; test what happens to records already written. The home page also describes self-healing Next Gen Zaps, labelled early access.
Workato's security page describes audit logs of executions, configuration changes and user actions, with 90-day configurable retention and longer retention through an external SIEM. Tray.ai describes logging of every workflow run, agent action and MCP tool call, but its pricing page ties the depth to the plan: log streaming to Splunk, Datadog or a SIEM is an Enterprise feature and an add-on for Team, and Pro and Team keep logs for 7 days by default (Team's retention is expandable, with a 30-day add-on). Neither vendor's pages describe automatic retries. That is a documentation question, not evidence that the feature is missing.
Billing units: one workflow, four different meters
n8n counts one execution as one complete workflow run, however many steps it has: 2.5k per month on Starter, 10k on Pro, 40k on Business (self-hosted only), with capped overage on Starter and Pro. Zapier counts every step in a Zap and every external connector call as tasks, and the rate varies with AI model tier, code runtime and connector type. All Zapier products draw on one shared pool, and reaching the limit switches you to pay-as-you-go unless you turn that off.
Workato uses a unified credit that covers tasks, API calls, rows processed and agent actions. Free gives 50K one-time credits; Pro is $100 a month, shown with 3.5k credits, while its FAQ describes recurring credits from 2,500 up to 50,000 with add-ons. Reconcile the two figures before budgeting. Tray.ai bills in tasks with no per-user, per-connection or per-runtime fees, lists HCU as a second unit in its FAQ, and quotes every tier on request.
Editorial suggestion, not a vendor capability: take one real workflow with a trigger, a lookup, a write and a notification, and ask each vendor to state in writing how many billing units one successful run and one failed-then-retried run consume. No cost is computed here because the units cannot be converted into one another.
What named customers run on these platforms
Zapier's SisterLove case describes a one-person marketing team: filling four columns in a Google Sheet triggers a Zap that has ChatGPT draft blog, email, SMS and social content, and the vendor reports more than 24 business days saved in nine months. Veo's story moves Facebook and Instagram lead data into ActiveCampaign and Pipedrive in real time; the vendor reports email journeys rising from 10 to 50.
Atlassian's Workato story concerns an ERP move from NetSuite to Oracle Fusion Cloud in nine months, with Workato carrying 800+ business process changes and 50+ integrations; Workato reports about 40,000 finance hours saved. At Tray.ai, Apollo built a Slack IT agent working across Okta, Kandji and ServiceNow that escalates to a human when it cannot resolve a request, and Tray.ai reports 40% ticket deflection in two weeks. Life360 routes requests to sub-agents and uses Tray workflows to audit and archive Slack channels.
These are vendor-published stories with different metrics, not a benchmark. For n8n, Dell and Cummins appear on its pages, but the sources read here do not describe their workflows.
A failing-workflow test for choosing the final two
Editorial recommendation: build the same workflow in each finalist, then make the write step fail halfway. Observe whether the run retries by itself, whether a record is created twice, how fast you can find the failed run in the logs, and what the vendor counts on its meter. Run it on the plan you would actually buy, since log retention, SSO and environments change by tier.
Method and limits
2026-11-20
This dated snapshot counts distinct companies in public customer evidence. CONFIRMED means a case study, named testimonial or press source; DECLARED means the company itself told usedby; OBSERVED means a vendor listing; SIGNAL means a job post. All four count, but their strength differs. Removed and unlevelled lines are excluded. Examples are a capped selection verified within 90 days. Vendor facts are sourced records checked within 60 days. Coverage is uneven and missing evidence is unknown. These figures measure neither market share nor product performance; capture dates are not adoption dates.