Automate large service upgrades with self-checking agents
Brex used Cursor agents to let one engineer update 180 microservices to a newer Micronaut release in roughly a month.
- What they did
- Brex assigned the project to two Cursor agents running on self-hosted infrastructure. A Linear ticket activated a generation step that paired scripts with model reasoning, compiled the code on a local setup, corrected its own errors, and produced a draft pull request. A GitHub label then launched a validation step that sent the change to dev, executed integration tests, and attempted fixes, with eleven small skills available to handle specific failure patterns.
- How it went
- One engineer completed the work across 180 microservices in about a month, an efficiency Brex estimated at 32x per person. Per-service cost dropped by around 90%. A handful of pull requests still looked good in tests but later led to subtle production bugs.
- Tips
- A single long prompt from a laptop did not perform well. Moving the agents to the cloud, breaking the logic into smaller skills, and stopping after three repair attempts made the system more dependable, but people still needed to review the output because test coverage missed some problems.
Try it yourself with Cursor
Set up a pair of self-correcting agents for [list of services or repos] that are being moved from [current framework version] to [target framework version]. When I open a [ticket type], have the first agent make the change, run a local build, diagnose failures, and repair them with targeted skills. Stop before any pull request is created and ask me to approve the result. After approval, have a second agent deploy to [dev environment], run integration tests, and try up to three fixes for any failures before it reports back.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Let a background agent watch tasks and open PRs
A team keeps a server-based agent running so Cursor can monitor work and turn finished tasks into pull requests.
Have agents co-build a reference system with role splits
A solo founder divided an AI-assisted build into roles and shipped a 120-model mental-models framework with an MCP server.
Let Cursor draft a project app in a 24-hour loop
A developer let Cursor work for a full day on a project tracker, and it finished with a usable but unfinished result.
Use Cursor to spot prototype gaps and draft a build plan
A founder used Cursor to inspect a trip-planning app, surface missing pieces, sketch screens, and then build a feature with parallel agents.
Turn saved bookmarks into quick demo builds
A daily bookmark check can feed Cursor a new tech idea and produce a prototype after approval.
Use Cursor to run Terraform through MCP
A developer used Cursor with tfmcp to plan and apply Postgres Terraform changes, then fell back to the terminal when permissions blocked the first attempt.