Getting Started
Quickstart
Turn your first idea into a working product
This quickstart will cover:
Congratulations! You’ve completed the quickstart.
Now you can start building your product with BrainGrid!
- Creating an account
- Connecting GitHub and analyzing your codebase
- Installing the BrainGrid MCP
- Writing your first requirement document
- Breaking down your requirement into tasks
- Building it with your favorite AI coding tool
- Automating acceptance criteria reviews
1
Create your Account
Go to https://app.braingrid.ai and create an account.Once you’ve got your invite, create an account using your email, google, or github account.Give your organization a name and pick a username.
We’re in private beta, so you’ll need to join the waitlist to get access.
We’ll send you an invite as soon as we’re ready for you.
2
Connect GitHub and analyze your codebase
Click on the “Connect GitHub” button to connect your GitHub account. BrainGrid needs to analyze your codebase to help you create code-aware tasks and requirements.
Choose repository access.
We recommend selecting only the repositories you plan to use with BrainGrid.Then, select the first repository you want to analyze. You can add more repositories later.
Click on the “Analyze now” button to analyze the repository. BrainGrid will start analyzing your codebase.Go to the Repositories page and see the documentation generated. These documents provide context for the BrainGrid agent to understand your codebase.
If the analysis is still running, you can get started getting the MCP set up while it’s done.
You must have Admin or Owner permissions to install the BrainGrid GitHub App on a repository. Learn more about how to install the BrainGrid GitHub App.
This may take a few minutes depending on the size of your codebase.
3
Install the BrainGrid MCP
Now let’s connect BrainGrid to your AI coding assistant using the Model Context Protocol (MCP).
Add BrainGrid to your Claude Code configuration:Then restart your Claude Code session. You’ll see that the BrainGrid MCP is available but needs authentication.
Enter Then you’ll see “BrainGrid disconnected - enter to login”
Hit “Enter”, this will open a browser window to login to BrainGrid. Once you log in, you can approve the BrainGrid MCP to connect to your account.
Once the authentication is complete, you’ll see “Authentication successful. Connected to BrainGrid.” and you can start using the BrainGrid MCP.
You can now use the BrainGrid MCP to create requirements and tasks.
/mcp to manage your MCP servers:Need more help? Check our detailed MCP installation guide.
4
Write your first requirement document
Next, let’s write your first requirement document.Click on the “Create requirement” button.
For the purposes of this quickstart, we are going to write a requirement from scratch. You can also refine an existing requirement or break it down into tasks.
Select the repository that you analyzed in step 2. If you need to build this requirement in a different repository, go to the Repositories page and click “Analyze now” to analyze the repository.
Once you’ve selected the repository, you can describe what you want to build. For this Quickstart, we are going to use the example of “Need to support two-factor authentication”.Feel free to jot down any additional context that you think is important.
The agent will start thinking and analyzing the codebase. It can come back with some clarifying questions to help it understand the requirement better.
Once the agent has enough clarity, it will start writing the requirement document.
At this point, you can review the requirement document and make any changes you need. You can ask the agent to make changes, clarify things, add or remove details.Once the requirement reflects what you need, you can move on to the following step to break it down into implementation tasks.
5
Break down your requirement into tasks
To break it down into implementation tasks, simply go to the “Tasks” tab and click “Break it down into tasks”.This will instruct the agent to break the requirement into atomic and verifiable tasks for your AI coding agent.
The agent will proceed to look at the last details of the implementation, and break this requirement into properly scoped tasks for your AI coding agent.
Once the tasks are created, you can review them and make any changes you need. You can ask the agent to make changes, clarify things, add or remove details.Tasks provide a very detailed prompt for your AI coding agent to follow.
Once you are ready to start building, go to the next step to build it with your favorite AI coding tool.
6
Build it with your favorite AI coding tool
You can simply copy the tasks and paste them into your AI coding tool like Cursor or Claude Code.You can also use the MCP and tell it to:
The AI coding tool, in this example Claude Code, will use the MCP to fetch the requirement and all the tasks.
You can tell it to start building, or to build only certain tasks if you want to stop at a certain point to review.
7
Automating acceptance criteria reviews
Lastly and most importantly, you can automate the acceptance criteria reviews.You can tell it to review REQ-218 so it reviews the implementation against the requirements.