Scale workloads, not headcount
Volume climbs and your workloads scale automatically. Nobody sizes workers, Redis or a database to keep up.
volume 9×
headcount +0
people added to keep up · none
Moving from n8n / Scale and reliability
The workflow works.
Scaling it is now your job.
Self-hosted, every step up in volume is sizing work your team does: workers, Redis, Postgres, retention, and the spikes in between. Tray scales automatically.
Thanks — someone from our team will be in touch soon.
Scale with confidence on Tray
n8n runs at volume for plenty of teams. Who does the sizing work, and who keeps doing it, is the question.
Spikes are absorbed
Traffic climbs and the platform takes it. Nobody provisions workers ahead of a busy day, and nobody re-architects after one.
capacity
peak
provisioned ahead nothing
Growth is not a re-architecture
Queue mode, retention and pruning are ours to run, so the next step up in volume is not a project.
10K
100K
1M
10M
migration projects 0
See across every run
What is eating resource, what is queuing behind something else, what finished short. Not assembled from other tools.
run resource q time
enrich-accounts 2 2.1s
sync-orders 0 0.9s
nightly-export 5 queued
score-leads 0 short
one place · not assembled
Patched automatically. Grows elastically.
Every advisory and every upgrade lands on a runtime we run and get audited on, however fast you are growing.
CVE v4.19 CVE
growth pauses patching never
Proven at volume
1T+ processes run per year on the platform, across every customer on it.
processes / year 1T+
Why enterprises choose Tray
What running it yourself takes
All of it from n8n's own documentation, and none of it a defect. It is simply the work.
- Memory. Large payloads are held in memory, and the documented fixes are batching, streaming and a higher memory limit, per workflow.
- Merge defaults. Unpaired items are left out by default, and the execution still succeeds.
- Queue mode. Scaling out moves execution data through a database, which n8n documents as an architectural constraint rather than a tuning problem.
- Operations. Retention, idempotency and visibility across thousands of runs are yours to design and keep configured.
Industry recognized
3× Visionary Gartner Magic Quadrant for iPaaS, 2024, 2025 and 2026 
7× Leader Nucleus Research iPaaS Value Matrix, seven consecutive years Pioneer Gartner Emerging Market Quadrant for No-Code Agent Builders, 2026 14 Hype Cycles Gartner inclusions in 2026, including Agentic AI and Agentic Automation On Tray, your workloads scale automatically
Nothing on that list becomes a ticket. Volume climbs and the platform absorbs it, with 100% workflow execution uptime over the trailing 90 days, measured publicly on status.tray.ai as of August 2026.
- Sizing and capacity are the platform’s job, so growth in volume is not a re-architecting project handed to your team.
- Large payloads stream and batch automatically, so volume moves through without anyone tuning a memory limit or splitting a workflow by hand.
- Observability across executions, not one at a time: what is consuming resource, what is queuing, and what finished short.
- Retention and pruning are operated for you, so an execution store filling up is not a disk nobody was watching.
Audited and certified
The controls are already on
SOC 1
audited annually
SOC 2 Type 2
with penetration testing
HIPAA
PHI handling
GDPR
EU data residency
CCPA
US data residency
700+
connectors managed
across integration, automation, and agents
1T+
processes run per year
on the platform
100%
workflow execution uptime
over the trailing 90 days
US, EU, and APAC data residency · Annual penetration testing · Uptime measured on status.tray.ai and publicly checkable there. As of August 2026.
See the trust centerKeep reading
BlogRisk and patchingWhen the next n8n CVE drops, who patches it?BlogOperational overheadWhat it costs to run n8n yourselfBlogScale and reliabilityWhy n8n drops records without an error, and what you own at volumeBlogGovernance and complianceWho owns your SOC 2 report when you self-host n8n?BlogMigrationIntroducing the n8n to Tray migration kit
More on moving from n8n:Risk and patchingOperational overheadGovernance and complianceMigrationStart from the top
Contents
01 02
02 04
03 06
04 08
05 10
tray.ai · 12 pages
The guide · 12 pages
Five signs you've outgrown n8n
Five signs, what each one costs, and where your setup stands.
Get the guideWhy enterprises love Tray
“Tray is the holy grail for software development. It has allowed us to build a single master workflow that is infinitely configurable and scalable without cumbersome code.”
Frequently asked questions
Does the n8n merge node drop items? +
By default it leaves unpaired items out, and this is documented behaviour rather than a defect. n8n’s own docs give the example: five items in Input 1 against ten in Input 2 processes five. It is deterministic and it happens at any volume, so it is not something you can load-test your way out of. It surprises people because the execution succeeds, and nothing tells you the other five went nowhere.
Can n8n handle high throughput? +
Yes, and n8n publishes a figure of 200 executions per second. Capability is not the question. Who does the sizing work is: queue mode, worker counts, Redis and Postgres tuning, memory limits, and execution-data retention are all yours to configure and keep configured.
Why does n8n run out of memory on large data? +
Because processing a large payload means holding it in memory, and n8n documents this along with the remedies: split the data into batches, stream where possible, and raise the Node process memory limit. All three are real fixes. All three are your team’s to implement and then maintain as data volumes shift.
What is the execution data constraint in queue mode? +
In queue mode, execution data moves between components through the database rather than staying in a single process, and n8n’s own forum describes this as an architectural constraint rather than a tuning problem. That distinction matters: a tuning problem has a setting, and an architectural constraint has a workaround you build and own.
How do I stop the execution database growing without limit? +
Set a retention policy and prune, which n8n documents. It works. It is also one more piece of operational configuration to get right, monitor, and revisit, and if nobody owns it the first symptom is usually an outage that turns out to be a disk.
How much sizing is your team doing?
Let's discuss how to scale without the work.
Thanks — someone from our team will be in touch soon.
GARTNER is a registered trademark and service mark, HYPE CYCLE and MAGIC QUADRANT are registered trademarks of Gartner, Inc. and/or its affiliates in the U.S. and internationally and are used herein with permission. All rights reserved. Gartner does not endorse any vendor, product or service depicted in its research publications and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s Research & Advisory organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.