Helix Full Tutorial - From Sign Up to Pull Request in One Flow

A complete walkthrough of Helix, covering every step from creating your account to merging AI-generated code into your repository.

What this tutorial covers

Signing up and onboarding (organization setup, subscription activation, AI provider config)
Creating projects and connecting GitHub repositories
Code intelligence: browsing the auto-generated wiki, semantic search, keyword search, and the changelog
Writing spec tasks and running the AI planning agent
Reviewing specs, giving feedback, and kicking off implementation
Using split-screen mode to run multiple tasks in parallel
Opening pull requests directly from Helix
Handling merge conflicts by sending them back to the agent

When you first visit Helix, you will see the login screen. Sign in with Google, enter your name and email, and your account will be submitted to the waitlist. Once approved, the onboarding flow walks you through creating an organization, activating a business subscription, and adding an AI provider. The built-in Helix providers work out of the box on the SaaS, or you can configure your own if you are self-hosting.

Creating a project and connecting repositories

During onboarding you can create your first project and connect a GitHub repository. The simplest approach is to follow the OAuth flow, which sets up all the right permissions. You can add more repositories later from the repositories tab within your project using the "Connect & Browse" button.

Code intelligence

Once a repository is indexed, you unlock several code intelligence tools:

Wiki — an auto-generated wiki for your codebase
Semantic search — ask a question or describe what you are looking for in natural language
Keyword search — find specific classes, functions, or identifiers
Grep and list files — file-level operations for browsing the repository
Read file — inspect individual files, exactly as the agent sees them
Changelog — an automated commit-by-commit changelog

These are the same tools exposed to agents via MCP, so the search interface doubles as a way to verify what the agent can see.

Spec tasks and planning

Spec tasks live inside a project. Create one by clicking "New Task" and describing the change you want. When you start planning, the agent enters a planning loop and pushes specifications to a dedicated Helix branch in your repository.

You can review the plan, approve it, or highlight text you are unhappy with and submit feedback. The agent will re-enter the planning loop and produce an updated specification.

Split-screen mode

One of the most powerful features is split-screen mode. You can view multiple agent desktops side by side, each working on a different spec task in parallel. Panels can be maximized, minimized, added, or removed.

Implementation and pull requests

Once you approve a plan, the agent moves into implementation. You can follow along live in the desktop view or context-switch to other tasks while it works.

When implementation finishes, click "Open PR" to create a pull request in GitHub. From there, follow your normal review process — run CI, request reviews, check preview deployments — and merge when ready.

Handling merge conflicts

If GitHub reports merge conflicts, go back to the agent and tell it to fix them. The agent will rebase and push back to your PR branch. Once CI passes again, you can merge as usual.