Developer Setup

Before you begin, ask your admin to complete Admin Setup for your repository. You need access to that repository and its issue tracker, Git, and a supported coding agent: Claude Code, Codex, or GitHub Copilot CLI.

You'll run two kinds of commands:

• Terminal commands, such as codelantern setup, run in a shell such as Terminal or PowerShell.
• Skills, such as /cl-dev:cl-init, run inside your coding agent's chat. Each agent invokes them in its own form (see step 3).

1. Join your organization

Accept your invitation and sign in to the CodeLantern portal with the invited email address. Check that the correct organization is selected. If the invitation has expired or you cannot see the organization, ask your admin to check your membership.

Open Personal settings → Connected accounts and connect your own accounts for the providers your repository uses. These connections are specific to the selected organization. Your admin's integration connects the organization; your personal connection lets CodeLantern act with your provider permissions. See Integrations for connection and permission troubleshooting.

2. Install the CLI

The public distribution is codelantern-ai/codelantern-cli. It supports macOS, Linux, and Windows on 64-bit Intel/AMD (x64) and Arm64 processors.

macOS or Linux

Run in a terminal:

curl -fsSL https://raw.githubusercontent.com/codelantern-ai/codelantern-cli/main/install.sh -o install-codelantern.sh
bash install-codelantern.sh

Windows

Run in PowerShell:

Invoke-WebRequest https://raw.githubusercontent.com/codelantern-ai/codelantern-cli/main/install.ps1 -OutFile install-codelantern.ps1
powershell -ExecutionPolicy Bypass -File .\install-codelantern.ps1

The installer verifies the download's checksum and prints any PATH changes you need. Follow that guidance, open a new terminal if needed, and check:

codelantern --version

It prints the installed version, for example codelantern version 0.1.7.

If the command is not found, check that the installation directory is on your PATH. The default is ~/.local/bin on macOS and Linux, or %LOCALAPPDATA%\Programs\codelantern on Windows.

The CLI isn't code-signed yet, so the first run may show a macOS "unidentified developer" warning or a Windows SmartScreen prompt. Allow it once. For other installation or PATH problems, open an issue in the codelantern-cli repository.

3. Sign in and set up your coding agent

codelantern login
codelantern setup

Complete sign-in in your browser. Setup uses your organization's plugin marketplace, detects supported coding agents, and asks you to confirm its proposed changes. It installs the cl-dev plugin, which provides CodeLantern's development skills, for your user account.

To configure only one coding agent, use the corresponding command:

Follow setup's instructions for your agent. Restart the agent after installation; for a desktop app, quit and reopen the whole app. Check that the cl-dev skills are available by invoking one in your agent's form. doctor checks your CodeLantern setup; before the repository is initialized, it replies that the project isn't initialized yet, which still shows the skill is installed.

The rest of these docs show skills in Claude Code form, such as /cl-dev:spec. In another agent, use that agent's form of the same skill name. If setup reports no organization marketplace, ask your admin to check the organization settings. If it reports an expired sign-in, run codelantern login and retry setup.

Released CLI versions check for binary updates automatically. The CLI and cl-dev plugin are separate: use your coding agent's plugin controls to manage plugin updates.

4. Initialize the repository

You initialize a repository once. If .codelantern/config.json is already committed, a teammate has done it and you can skip to the next step. Otherwise, open your local repository in your coding agent and invoke the cl-init skill in your agent's form (see step 3), for example:

/cl-dev:cl-init

The skill prepares the repository's configuration, project instructions, and starter knowledge base, and asks about optional repository automation. It shows a summary when it finishes and doesn't commit anything. Review its changes, then commit them and merge them into your default branch. See cl-init for its prerequisites, what it creates, the choices it asks about, and safe re-runs.

5. Prepare the knowledge base

Skills that plan, build, and review work read the knowledge base for context, so fill it before your first feature. This is recommended rather than required, and it's done once per repository. If .codelantern/knowledge-base/project.md and conventions.md already describe the project, a teammate has done this and you can go to your first task. Otherwise, run these in order:

/cl-dev:discover

discover scans the codebase and proposes what it finds: what the project is and how it's structured, the conventions the code follows, and reusable UI components if the repository has a user interface. It asks before writing.

/cl-dev:extract

extract interviews someone who knows the project about the reasoning behind decisions, the team's vocabulary, and conventions nobody wrote down, which a code scan can't find. Run it with that person at the keyboard. It checks with you before saving each entry.

If the repository is new or has little code, start with extract and run discover once the codebase takes shape. Commit the knowledge-base changes like any other change. The knowledge base keeps growing as you work: /cl-dev:consolidate adds what each feature teaches.

Ready for your first task

You should now have an active organization membership, personal provider connections, cl-dev available in your coding agent, an initialized repository, and a knowledge base with context about your project. If the agent reports that the repository is missing or incomplete, ask your admin to check its repository configuration.