Getting started

Before a workflow can run, Expecto Centum needs a few things in place: a coding agent to do the work, a repository to do it in, and — when your tickets live in a tracker — a ticket source with a filter that says which tickets to pull. This manuscript walks through each, in the order you set them up.

On first launch, the app's setup wizard covers the first steps: it finds the coding agents on your computer, names your first workspace, and asks whether you're starting a new project or continuing existing code. Everything it sets up can be changed later in Config, behind the gear at the top left of the dashboard.

I Install a coding agent

Expecto Centum doesn't bring a model of its own. Every stage that thinks — triaging a ticket, planning its tests, writing the code, reviewing it — runs through a command-line coding agent installed on your computer and signed in with your own account. It works with:

• Claude Code, from Anthropic
• Codex CLI, from OpenAI
• Gemini CLI, from Google
• OpenCode — open source, using whichever providers you've set it up with
• A custom command — any other command-line agent, described by the command that runs it

Install at least one and sign it in the way its own instructions describe — a subscription login and an API key both work. Then check that it runs from a terminal: the app looks for agents wherever a new terminal window finds programs, so an agent you can run there is found even when you open the app from the Dock or the Start menu.

II Register it with the app

Registering gives Expecto Centum permission to run that one program. There are two ways:

• On first launch, the wizard asks Which coding agents should Expecto Centum use? and lists the agents it found, each checked. Press Add and continue. One installed somewhere the app didn't look can be added with Find an agent….
• At any time: Config → Connections → CLI agents → Register CLI agent. Choose the agent and press Check & register — the app makes sure the program starts before it saves it.

Each stage then runs on the agent chosen for it in Config → Workflow, where you can also pin a model per stage; left alone, a stage uses your agent's own default model. Before every run the app starts each agent the run needs, so a missing or broken install stops the run before any ticket is touched and names the agent, rather than failing halfway through.

III Add a repository

A workflow works inside a git repository on your computer.

• Existing code — the folder has to be a git repository checked out on a branch. Add it under Config → General → Local repositories, or pick it when the wizard asks and you choose Continue existing code. Each repository has a Default branch: the run starts its work from that branch and merges finished work back into it.
• A new project — there's nothing to prepare. Greenfield's Specifications stage makes the repository for you: press Create, then Create new repo.

You can add as many repositories as you like. On Brownfield, each ticket is matched to the repository it belongs to.

IV Connect a ticket source

To pull tickets from a tracker:

1. Open Config → Connections → Ticket sources and press Connect ticket source.
2. Choose your tracker — Asana, ClickUp, GitHub Issues, GitLab Issues, Jira or Linear — and press Connect.
3. Approve the connection in your browser. There are no tokens to copy; once you approve, the connection is listed under Ticket sources.

Connect a ticket source covers each step and what to do when one goes wrong. Asana needs an app of your own first — see Connect Asana.

V Add a filter

A connection on its own scans nothing. A filter says which of its tickets to pull.

1. Open Config → Workflow → Ticket Scan. Every connected tracker is listed under Ticket sources.
2. Press Add filter on the tracker and choose what to scan — a Jira project, a Linear team, an Asana project or your My Tasks, a ClickUp list, a GitHub or GitLab repository — and, if you like, narrow it by status, label, assignee or a search.
3. Keep the filter checked. A scan pulls the tickets of every checked filter at once, across all your trackers; an unchecked filter is kept for later and scanned by nothing. A new filter starts checked.

Checks are kept per workflow and per workspace, so the same filter can be scanned by Brownfield and left out of Greenfield.

VI Connect a Git host

This step is optional. A run does all of its work on your computer: it writes the tests and the code, reviews and documents them, and commits everything to your local repository. Pushing that work and opening pull requests needs a plan from Tenfold and a Git host connection — Add GitHub, Add GitLab or Add Bitbucket under Config → Connections. GitLab and Bitbucket need a step on their side first: see Connect GitLab and Connect Bitbucket. The repository also needs a remote on that host, which you can add from its row in Local repositories.

VII Check the dashboard

The dashboard tells you when something is missing. A stage that can't run yet wears a warning badge, and the buttons that would run into it stay disabled, with the reason in their tooltip:

• No CLI agent registered — register one (step II).
• Missing CLI agent — a stage is set to an agent that is no longer registered. Register it again, or choose another for the stage.
• Nothing to scan — no tracker connected, no filter added, or no filter checked (steps IV and V).

Click a badge to go straight to the setting that fixes it. With no badges left, press Scan — or Create on Specifications — then Start. Built-in workflows explains what each stage does and which workflow to run.