Minimal Gerrit Workflow with Docker and Git for Agentic Coding

I wanted to experiment with a workflow that lets an AI write and push code to version control, where I can review it, and the AI reads those code reviews and acts on them. I heard a long time ago about Gerrit, a code review tool built on top of Git. More specifically, Gerrit works with reviews on a commit basis, which I don’t have much experience with, but is an interesting concept nonetheless. I would say that as a takeaway, I didn’t find it particularly helpful for a smoother workflow with AI agents, but still, it was nice to play with it.

Start Gerrit in Docker

docker run -ti -p 8080:8080 -p 29418:29418 gerritcodereview/gerrit

Runs Gerrit in a Docker container, exposing the web UI on port 8080 and SSH on 29418.

Add SSH key to Gerrit via API

curl -u admin:secret \
     -X POST \
     -d '{"ssh_public_key": "'"$(cat ~/.ssh/id_rsa.pub)"'"}' \
     http://localhost:8080/a/accounts/admin/sshkeys

Uploads your public SSH key to the admin account using Gerrit’s REST API. Unfortunately this does not work, so you need to add them manually on http://localhost:8080/#/settings/ssh-keys

Create a new project in Gerrit

ssh -p 29418 admin@localhost gerrit create-project vibecode --branch main

Creates a new project named “vibecode” with a default main branch.

Clone the repository over SSH

git clone ssh://admin@localhost:29418/vibecode

Clones the newly created project from the Gerrit server.

Install Gerrit’s commit-msg hook

curl -Lo `git rev-parse --git-dir`/hooks/commit-msg http://localhost:8080/tools/hooks/commit-msg && chmod +x `git rev-parse --git-dir`/hooks/commit-msg

Downloads and installs the commit-msg hook required to add a Change-Id to commit messages.

Push a commit for review

git push origin main:refs/for/main

Pushes the current branch to Gerrit for code review.

Query a change and its comments

ssh -p 29418 admin@localhost gerrit query --comments --current-patch-set 27eac4ed28b66d45e0ad30e795dda14ba70925d3

From here on, you can pass the output to a coding agent and let it figure out the improvements that need to be done on its own.