Agent pool

What it is

An agent pool is a named collection of agents that execute jobs. In Azure Pipelines you assign a job to a pool with pool:; the job then runs on any available agent in that pool.

Why it exists

Pipelines need compute to execute steps. Instead of pinning each job to one machine, a pool gives elasticity: many agents, parallelism, and choice between Microsoft-managed and your own hardware.

Key ideas

• Microsoft-hosted pools — pre-provisioned, managed VMs; named by OS (ubuntu-latest, windows-latest, macos-latest). Zero setup; billing by hosted minutes.
• Self-hosted pools — you register your own agents into a pool (org/project level). More control, no per-minute cost, but you maintain them.
• Organization vs project pools — org-level pools are shared; project pools are scoped to one project.
• Parallelism — more agents in a pool ⇒ more concurrent jobs.
• GitHub Actions equivalent: runners (GitHub-hosted or self-hosted) chosen via runs-on:.

Exam notes

• Syntax: pool: vmImage: 'ubuntu-latest' (Microsoft-hosted) vs pool: mySelfHostedPool (self-hosted).
• Demand/demands: — self-hosted agents can declare capabilities (e.g. dotnet), letting jobs pick capable agents.
• Hosted pool names/vmImage values are common exam items.

Related

self-hosted-agent · azure-pipelines · jobs-steps

📘 Source: Microsoft Learn — Agent Pool