Azure Pipelines · Rate Limits

Microsoft Azure Pipelines Rate Limits

Azure DevOps Services (which hosts Azure Pipelines) applies a per-user / per-identity Throughput Service Threshold (TSTU / RCU) sliding-window rate limit. Each REST API call consumes a number of TSTUs based on cost; each user has 200 TSTUs per 5-minute sliding window. Throttled requests return HTTP 429 with Retry-After. Concurrency is capped not by API rate but by purchased parallel jobs (default 25 for Microsoft-hosted on new orgs, raise via support). Per-user TSTU limits cannot be raised — clients must throttle.

Microsoft Azure Pipelines Rate Limits is the machine-readable rate-limit profile for Azure Pipelines on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 5 rate-limit definitions, measuring tstu_per_5_minutes, concurrent_jobs, and minutes_per_run.

The profile also includes 5 backoff/retry policies defined and response codes documented for throttled and delayedNotBlocked.

Tagged areas include Automation, Build, CI/CD, Deployment, and DevOps.

5 Limits Throttle: 429
AutomationBuildCI/CDDeploymentDevOpsPipelinesRate Limiting

Limits

Per-user TSTU sliding window user
tstu_per_5_minutes · minute
200
Throughput measured in Throughput Units (TSTUs). 200 per 5-minute sliding window per user. Heavy queries cost more TSTUs; light reads can cost <1.
Microsoft-hosted parallel job concurrency organization
concurrent_jobs
25
New organizations are capped at 25 Microsoft-hosted parallel jobs until a support ticket raises the limit; subject to regional capacity.
Microsoft-hosted job duration (paid) pipeline-run
minutes_per_run · minute
360
Each paid Microsoft-hosted job has a 6-hour per-run timeout.
Microsoft-hosted job duration (free private) pipeline-run
minutes_per_run · minute
60
Free private projects cap individual runs at 60 minutes.
Self-hosted job duration pipeline-run
minutes_per_run
-1
No per-run timeout on self-hosted parallel jobs.

Policies

TSTU cost weighting
API calls cost differing TSTU values based on database/CPU expense. Inexpensive reads may cost <1 TSTU; expensive list/query operations may cost 5+.
Delay before block
Before returning 429, the service may inject a delay (X-RateLimit-Delay) header to slow the client without blocking. Persistent abuse upgrades to 429.
Spread automation across identities
Build agents under PAT or service principal share the same per-identity quota; use multiple service principals to fan out automation.
Exponential backoff
Honor Retry-After. Implement exponential backoff with jitter for 429 responses.
No support increase for TSTU
Per-user TSTU limits cannot be raised by support. Design clients to respect headers and stay under limit.

Sources