Projects serve as containers that organize your AI agent runs around a specific task, feature, or initiative. By grouping related runs together, you can:

  • Maintain context continuity across multiple agent interactions
  • Share common documentation and requirements with all agents
  • Track progress on complex, multi-step tasks
  • Collaborate effectively with team members on larger initiatives (coming soon)

Creating Projects

There are two primary ways to create a project:

1. Create from Projects Page

Navigate to the projects page and click “New Project” to create a project from scratch.

2. Create from Design Document

Generate a project directly from an existing design document or specification.

Why Use Projects?

Context Window Limitations

All AI agents have limited context windows, meaning they can only process and remember a finite amount of information before their performance degrades. This limitation becomes critical when working on:

  • Large codebases with complex architectures
  • Multi-step feature implementations
  • Tasks requiring deep domain knowledge
  • Long-running development initiatives

The Solution: Shared Context

Projects solve this by providing a shared context document that all agents can reference. This approach:

  • Eliminates context loss between agent interactions
  • Ensures consistency across different implementation phases
  • Reduces repetitive explanations of project requirements
  • Improves agent performance by providing relevant background

Here’s a proven workflow for maximizing project effectiveness:

  1. Deploy a research agent to analyze your codebase and understand the current architecture. Generate a comprehensive design document that outlines requirements, constraints, and implementation strategy
  2. Create a project from your design document with clear naming.
  3. Trigger specialized agents for specific implementation tasks. Leverage the shared context to maintain consistency across agents

Benefits

By using projects effectively, you’ll:

  • Save time by eliminating manual context copying between chats
  • Organize your work with clear project boundaries and run history
  • Improve performance through consistent, well-informed agent interactions
  • Scale your development process for larger, more complex tasks
  • Maintain quality through shared understanding and documentation