Getting started with CodeCourier

This is the guide we wish we had handed every early beta user. It walks you through the exact path from a fresh CodeCourier account to a real pull request authored by an AI agent - sitting in your repo, with passing tests, ready to merge. No hand-waving, no "and then magic happens." Twenty minutes, start to finish.

1. Create your workspace

Sign in to CodeCourier with GitHub. The first time you log in we create a workspace named after your account. A workspace is a billing boundary, an access-control boundary, and a context boundary all at once - every Issue Session, Persona, and Sprint Chain you create lives inside it.

If you are evaluating CodeCourier for a team, do this with your work GitHub account, not a personal one. You can rename the workspace later and invite colleagues, but starting in the right account avoids re-doing OAuth flows.

2. Connect your first repository

From the dashboard, click Connect repo. Choose a repo small enough that you can spot-check the agent's work without ceremony - a personal project, an internal tool, or a hardened sandbox repo are all good first picks. Avoid your production monorepo for the first run.

CodeCourier asks for read + pull-request-create scopes by default. It never receives push access to main. Every change ships through a pull request you have to merge yourself.

The principle is simple: agents can propose anything, but a human merges. Your branch protection rules stay in charge.

3. Pick a starter persona

Personas are versioned agent configurations - model choice, thinking budget, tools, skills, and house style baked into one named profile. For your first session, pick Starter Engineer from the persona library. It is the safest default: a balanced model, moderate budget, and a conservative review philosophy.

You can customize personas later. The Design a Frontend Persona guide walks through that end to end.

4. Open your first Issue Session

Click New Issue Sessionon your connected repo. You will see a small form with three fields: title, description, and persona. For a confidence-building first run, paste a real bug from your tracker or pick something deliberately small - "the README still references the old CLI flag" works perfectly.

title:       Fix outdated CLI flag in README
description: README still mentions --legacy-mode which was removed in v2.0.
             Update all references to use --strict instead.
persona:     Starter Engineer

Hit Launch. A sandbox boots in under a second with your repo cloned, dependencies installed, and the persona loaded. You will see a live event stream - the agent reading files, drafting a plan, proposing edits, running tests, iterating until everything is green.

5. Review the pull request

When the agent is happy with its work it opens a pull request on your repo. Click through to GitHub. You will see a clean diff, a structured description of what was changed and why, and CI results from your existing checks.

Read the diff like you would for any junior engineer's first PR. Leave comments if something is off - the agent will read them and push a fix-up commit. Merge when you are happy.

6. Next steps

That is your baseline. From here the obvious next moves are:

• Try the Build a multi-step workflow guide to chain a planner, coder, reviewer, and shipper into one pipeline.
• Wire GitHub issue integration so new tickets become Issue Sessions automatically.
• Read the Operate CodeCourier in production guide before pointing agents at your real backlog.

You now have the muscle memory for every other guide in this hub. The rest is composition.