LightSprint Documentation

Getting Started

LightSprint helps teams align on requirements and execute them with AI coding agents. Here's how to get started:

1. Sign Up

Sign in with your GitHub account. LightSprint uses GitHub App authentication, so you'll need to authorize the app to access your repositories.

2. Create a Project

After signing in, create a new project and connect it to a GitHub repository. This gives LightSprint access to your codebase for context-aware task generation.

3. Start Collaborating

Use the AI-powered chat to align PMs, designers, and engineers on requirements before breaking them down into tasks.

Task Management

Tasks are automatically generated from aligned requirements, or you can create them manually.

Auto Task Generation

When requirements are aligned, click "Generate Tasks". LightSprint will:

• Analyze your codebase to understand existing patterns
• Break down the requirements into implementation-ready tasks
• Identify files that need to be created or modified
• Map dependencies between tasks

Kanban Board

Tasks are organized on a Kanban board with columns:

• Todo - Tasks ready to be worked on
• In Progress - Tasks currently being implemented
• In Review - Tasks with open PRs
• Done - Completed tasks

Task Details

Each task includes:

• Description and context
• Related files from your codebase
• Implementation todos (subtasks)
• Assigned AI agent (if applicable)
• Linked PR (when created)

Auto Task Tracking

LightSprint automatically updates task status based on your git commits. When you commit code that matches a task's todos, the task moves to the appropriate column.

AI Agents

LightSprint integrates with multiple AI coding agents to execute tasks automatically.

Supported Agents

• Claude - Best for complex reasoning and code review
• Cursor - IDE-integrated coding with full context
• Grok - Fast task execution

Launching an Agent

To launch an AI agent on a task:

1. Open a task from the Kanban board
2. Click "Launch Agent"
3. Select which agent to use (or let LightSprint choose automatically)
4. The agent will read the task context and start implementing

Agent Execution

Agents work in isolated branches and:

• Read the full task context (description, todos, related files)
• Create or modify files as needed
• Open a pull request when done
• Link the PR back to the original task

Parallel Execution

You can run multiple agents simultaneously on different tasks. LightSprint detects conflicts before they happen and ensures each agent works on isolated branches.

GitHub Integration

LightSprint deeply integrates with GitHub to provide seamless workflow.

GitHub App

LightSprint uses a GitHub App for authentication and repository access. This means:

• Single OAuth flow for both auth and repo access
• Granular permissions (only what's needed)
• Works with any repository you have access to

Webhook Events

LightSprint listens to GitHub webhooks to:

• Track commits and update task status
• Monitor pull requests and link them to tasks
• Detect branch creation and updates

Commit Analysis

When you push commits, LightSprint analyzes them to:

• Match commits to task todos
• Update task status automatically
• Provide confidence scores for matches

Best Practices

Task Management

• Break large features into smaller, focused tasks
• Review auto-generated tasks and refine them if needed
• Use clear commit messages to help auto-tracking work better
• Link related tasks together

Working with AI Agents

• Provide clear task descriptions with context
• Review agent-generated code before merging
• Use agents for repetitive tasks and boilerplate
• Let agents handle implementation while you focus on architecture

Team Collaboration

• Align on requirements before coding
• Comment on tasks to provide additional context
• Review PRs linked to tasks
• Keep the Kanban board updated for visibility