System Limits & Quotas

Check the batch size, concurrency, timeout, payload, and retention ceilings for automation runs.

At a glanceVerified in app · Oct 30, 2026

Use it to

Look up execution timeouts, payload ceilings, loop concurrency, retry limits, and storage quotas

Find it

Billing and usage → Usage, or validate_workflow over MCP

Limits

10 MiB webhook body · 60 s held webhook · 13 d Wait / Human Review · 256 KB testing data

Common execution limits

Some limits are checked during configuration. Others apply to runtime payloads.

Batch processing (Run for each item)

Run for each item enforces batch caps alongside resource balances and execution capacity.

Limit	Free	Paid	Behavior
Items per batch	50	1,000	Max array length at fan-out. In Repeater, sets pass cap.
Parallel item runs	1	5	Simultaneous items per batch.
Retry failed items	No	Yes	Continues past failed items and retries them.

• Credits: Each item charges at the Step rate (Step Credits, Tokens & Storage). Cap lists with Limit.
• Multi-Step loops: Run for each item works on single action Steps. Use Repeater for AI Agent, routing Steps, Combine, Respond to Webhook, or Wait.

List and transformation ceilings

Mechanism	Ceiling	When exceeded
Filter, Sort, Limit, and Remove duplicates	100,000 items per pass (Sort allows 10 sort rows)	Fails the Step naming both counts.
In-field list transformations	10,000 items	Fails the Step; filter or page first.
In-field text splitting (| split)	10,000 non-empty pieces (trims spaces, skips empties)	Fails the Step; split large text.

Retries

Retry on fail (yours to configure)

Enable Retry on fail in a Step’s Test & Debug tab for transient errors.

Setting	Range	Default	Notes
Retry on fail	on / off	Off	Retries retryable errors (Error Handling & Retries).
Max retries	0–10	3	3 retries allow 4 attempts (10 retries allow 11 attempts).
Wait between tries	0–60 seconds	5 s	Base delay adjusted by jitter and Retry-After.

Terminal failures, interrupted calls, and isolated Step tests skip auto-retries. In Run for each item, each item has its own retry budget.

Plan for repeated actions

An interrupted Step can still complete externally. Check the destination service before retrying writes, and use idempotency keys or stable IDs.

Timeouts

Scope	Limit	Notes
AI Agent active time	180s default, 600s max (min 10s)	Excludes approval waits; reserves up to 25s for the final answer.
Code editor run & memory	150s default; memory is environment-dependent	Filter large inputs upstream and confirm memory needs with your account team.
Code editor output	100,000 stdout / 20,000 stderr chars (256 KiB cap)	Keeps stream tails; omits result.output if final JSON is cut (truncated output).
Outbound HTTP request	30s default, 300s max	Use a Webhook callback for longer jobs.
Wait Step	13 days	Store state and poll with a Scheduler for longer waits.
Human Review expiry	13 days (1,123,200 seconds; or blank)	Routes to Expired if connected, or the error path otherwise.
Workflow Assistant Import JSON	512,000 characters	Split large exports (Migrating to Glow).

Payloads and storage

Scope	Limit	Notes
Payload to a Glow Webhook or App trigger URL	10 MiB default (1 MiB = 1,048,576 bytes)	Rejects oversized requests before starting a run. Send an ID or URL instead.
Configured webhook response body	1 MiB	Falls back to the standard 202 acknowledgement without dropping the delivery.
Outbound HTTP Request response	Up to 10 MiB per request	Request fewer fields or smaller pages.
Combined results of a loop	25 MB across all items	Fails Run for each item; return IDs only.
Single file upload & Team file storage	100 MB per file; plan quota in File Management	Rejects files over 100 MB; keep larger assets in external storage.

Data retention

Data	Default workspace policy
Run records and step-level input/output payloads	30 days after the run finishes

Each run keeps its starting policy; updates apply to new runs only. Retention excludes workspace files, AI Agent memory, and external services. Forward audit records to your own sink (Step-level Executions and Security & Compliance).

MSP impersonation sessions

Console sessions last 15 minutes, 30 minutes (default), or 60 minutes. Glow checks authorization on requests and open canvas connections. Withdrawn consent or reduced access closes active sessions (Governance & Impersonation).

Raising a limit

Paid plans unlock higher batch sizes, parallel item runs, and item retries (Billing & Usage). Contact your account team for Enterprise quotas.

Verify it works

Run

Open Billing and usage → Usage in your workspace or run validate_workflow over MCP

You should see

Your active plan's Step credit, Glow AI token, storage, and concurrency limits are displayed alongside current usage.

What’s Next?

• Design around failures with Error Handling.
• Diagnose a specific failure in Troubleshooting.