Cursor Builds: Environment Snapshots That Make Cloud Agents Start Instantly

Forum topic · 小凯 · 2026-08-14

Summary

Cursor introduced builds on August 13, a feature that keeps continuously prepared snapshots of cloud development environments so agents can start working within seconds. Previously, each cloud session required booting a machine, cloning the repository, and running installation scripts, which could take minutes for large codebases. Builds run hourly by default, capturing snapshots where the repository is already cloned, dependencies are installed, and setup scripts have completed; new sessions fork a live machine rather than restoring from cold storage. Cursor reports 10x faster environment startup and 3x faster time-to-first-token. Customer Faire, which runs over 2,000 automated agent jobs weekly, now boots its largest repositories in seconds. Since August 17, builds are enabled by default for all environments at no extra cost. Because builds rely on filesystem snapshots, install commands must cover anything pre-provisionable, and private registries should use team or environment keys, since user keys are injected only at agent startup. Cloud agents always start from the most recent successful build, so broken builds never disrupt running sessions.

Agents are only as capable as the environments they run in. Cursor made that point again in an August 13 blog post — and offered a concrete solution: builds.

The Speed Math

Until now, every cloud session launch went through the same three-stage ritual: boot the machine, clone the repository, run installation scripts. For large, complex codebases, just getting the environment ready could take minutes before the agent could do anything.

Builds work differently: Cursor continuously prepares ready-to-use copies of your development environment in the background, running a build once per hour by default. The snapshot captures a state where the repository is already cloned, dependencies are installed, and setup scripts have finished. When an agent starts, it forks a live machine instead of recovering cold from disk.

Internal results:

10x faster environment startup
3x faster time to first token

Customers see the same gains. Faire runs 2,000+ automated agent jobs per week; its most complex monorepos now spin up in seconds, and "a bad build can never take down the entire agent fleet."

As of August 17, builds are enabled by default for all new and existing environments, at no additional cost.

Boundaries

Builds rely on filesystem snapshots, so two things need manual attention:

1. Install commands should cover anything that can be pre-provisioned (such as dependencies). 2. Private registries requiring keys should use team or environment keys. User keys never enter snapshots and are injected only when the agent starts.

The start command still runs when you send your first prompt — it exists for services that must start fresh (for example, bringing up Docker).

Another reliability win: cloud agents always start from the most recent successful build. If a dependency upgrade breaks the install script, that build simply never takes effect. You debug in the background while new and existing sessions keep running normally.

Rethinking "Waiting"

Cursor has turned environment preparation from a per-session upfront tax into an always-on background pipeline. For agent engineering, the real bottleneck was never how smart the model is — it's whether the agent can get to work immediately. When startup drops from minutes to seconds and a broken environment no longer takes down the whole fleet, handing more engineering work to autonomous agents stops being a slogan and becomes an executable default.