Adopt Backstop

Try it out.

Install Backstop and see how the pieces work together.

1. Set up Backstop

Install the binary

Install a released build with Homebrew or Go. Or clone the repository and build it.

These commands cover macOS and Linux on amd64 and arm64. Windows is planned.

Homebrew:

brew install backstop-ai/tap/backstop

Go, pinned to a module version in the repository:

GOBIN=./.backstop-bin go install github.com/backstop-ai/backstop-core/cmd/backstop@v0.2.0

From a clone:

git clone https://github.com/backstop-ai/backstop-core.git
cd backstop-core
go build -o ./bin/backstop ./cmd/backstop

Initialize the repository

Backstop configuration lives with the code. Init writes the repository files later clones restore.

backstop init

2. Start with your existing code

Install a pack appropriate for the repository’s stack. Choose a pack.

Run:

backstop baseline

Review what Backstop finds. Do not fix anything yet.

3. Try it on a bug you know

Pick a known bug in the repository. You already understand it, so you can judge whether the issue and plan capture the work.

Bounded work uses issue → plan. This is the normal path for a bug, a follow-on, or a small fix.

4. Let the agent implement it

Assign the approved plan to an implementer agent. The agent executes the plan, runs required gates as it works, and fixes failures before proceeding.

Stop before merge.

A zero exit is the expected postcondition only when every blocking check passes. Keep the failing receipt when it does not.
backstop gate

Understand the enforcement loop

You’ve now used the whole model.

Use the complete workflow, or only the parts you need.