---
title: "Getting started"
description: "Set up Kai Code for a project in a few minutes: repositories, tools and customer replies."
url: https://docs.gleap.ai/en/articles/689-getting-started
collection: "Kai Code"
updated: 2026-11-01
language: en
---

# Getting started

Set up Kai Code for a project in a few minutes: repositories, tools and customer replies.

Kai Code is set up once per project, then every teammate can start sessions. You need the **Kai Code** feature in your plan; the dashboard offers an upgrade where it is missing.

## The setup wizard

Open AI agents → Kai Code → Settings to configure the project. New workspaces can also use the Onboarding setup chat in Kai Copilot. Connect the repositories Kai should work in, then review the settings below.

1.  Connect a repository
    
    Pick GitHub, Azure DevOps, GitLab or Bitbucket, authorize Gleap, and choose the repository Kai should work in. Set the default branch pull requests are opened against. This step is skipped when the project already has a repository. Details: [Repositories](https://gleap-developer-docs.gleap.help/en/articles/693-repositories).
    
2.  Give Kai Code its tools
    
    Optionally add [MCP servers](https://gleap-developer-docs.gleap.help/en/articles/695-mcp-servers): the Gleap preset, a Composio integration such as Linear or Notion, or a custom server. Kai Code works without any; you can skip this step.
    
3.  Decide how Kai Code talks to customers
    
    Choose what happens on a ticket once a fix is merged: a teammate verifies first, Kai replies automatically after a deployment delay, or nothing is sent. Details: [Customer replies](https://gleap-developer-docs.gleap.help/en/articles/696-customer-replies).
    

If you are using the setup wizard, choose Finish setup after reviewing its configuration. You can revisit the project under AI agents → Kai Code → Settings.

## Where sessions run: a personal choice

The first time you start a session, Kai Code asks **where Kai should run**. This is remembered per person, not per project, because it depends on what you have on your computer:

Kai Code Cloud

Always ready, nothing to install. Sessions run in Gleap's sandbox and are billed at API usage pricing from your organization's AI balance (see [Plans and AI pricing](https://gleap-developer-docs.gleap.help/en/articles/713-plans-and-ai-pricing)). The right choice for most teams and for teammates without a development setup.

Kai Code Bridge

Sessions run on your own computer or server with your own **Claude Code, Codex or Cursor** subscription and your local checkouts. One command to connect: see [Kai Code Bridge](https://gleap-developer-docs.gleap.help/en/articles/697-kai-code-bridge).

You can switch between the two for each new session with the **Run on** picker in the composer.

## Project defaults

Under **AI agents → Kai Code → Settings → General** an admin sets the defaults every session starts from:

-   **Default model** and reasoning effort for new sessions. Teammates can override both in the composer.
    
-   **Custom instructions** Kai follows in every session of the project, e.g. "Prefer functional React. Never edit files in `src/legacy`. Always run prettier before committing."
    
-   **Post results to the pull request**: add the verification result and a commit status to every pull request. Evidence such as screenshots stays private to Gleap.
    

## Next steps

-   [Sessions](https://gleap-developer-docs.gleap.help/en/articles/690-sessions): the composer, plan mode, verification and pull requests.
    
-   [Tickets and pull requests](https://gleap-developer-docs.gleap.help/en/articles/691-tickets-and-pull-requests): start Kai Code straight from a bug report.

## Related

- [Kai Code](https://docs.gleap.ai/en/articles/688-kai-code.md)
- [Sessions](https://docs.gleap.ai/en/articles/690-sessions.md)
- [Tickets and pull requests](https://docs.gleap.ai/en/articles/691-tickets-and-pull-requests.md)
- [Your own coding agent](https://docs.gleap.ai/en/articles/692-your-own-coding-agent.md)
- [Repositories](https://docs.gleap.ai/en/articles/693-repositories.md)