Skip to main content

Welcome to Bloom

Bloom is a multi-agent task orchestration system that enables AI-powered development through YAML-based task definitions and Claude Code agents working in parallel. Whether you're a solo developer or part of a cross-functional team, Bloom adapts to your workflow.

What is Bloom?​

Bloom bridges the gap between high-level project planning and AI-powered execution. It supports the entire development lifecycle—from requirements gathering to code generation—with built-in collaboration points for teams of any size.

  • Plan across repositories — Coordinate work spanning backend, frontend, and shared libraries in a single project
  • Explore your codebase — Use bloom enter to ask questions across all your repos without setting up a project
  • Plan thoroughly — Define requirements in PRDs with mockups, designs, and technical constraints
  • Collaborate naturally — PMs, designers, developers, and QA can each contribute at the right stages
  • Execute in parallel — Run multiple Claude Code agents simultaneously on different tasks
  • Stay in control — Human-in-the-loop design with checkpoints, question queues, and task oversight
  • Scale safely — Git worktrees isolate agent work, preventing conflicts
init workspace → clone repos → create project → refine PRD → plan → generate → run

For Teams and Solo Developers​

Teams benefit from structured collaboration points:

  • Product managers and designers own the PRD and add mockups or design links to the project folder
  • Developers, architects, and technical leads refine implementation plans together
  • QA engineers validate checkpoints at phase boundaries
  • DevOps and security teams can review and contribute to planning

Solo developers move fast:

  • Turn ideas into structured plans with AI assistance
  • Let multiple agents execute in parallel
  • Focus on reviewing and iterating rather than writing boilerplate

Key Features​

Multi-Repository Planning​

Unlike single-repo tools, Bloom coordinates work across multiple repositories in a single project. Build features spanning your backend, frontend, mobile app, and shared libraries—all with proper cross-repo dependency ordering.

Cross-Repo Exploration​

Use bloom enter to start an interactive Claude session with access to all your repositories. Ask questions about how systems connect, debug issues across services, or onboard to a new codebase—no project setup required.

Multi-Agent Orchestration​

Run multiple Claude Code agents in parallel, each working on isolated tasks in their own git worktrees. Automatic dependency resolution ensures tasks complete in the right order.

Human-in-the-Loop​

Agents can ask questions and wait for human responses. You stay in control of critical decisions while AI handles the implementation.

YAML-Based Tasks​

Define tasks with clear instructions, acceptance criteria, and dependencies. The schema is simple yet powerful enough for complex projects.

Git Worktree Isolation​

Each agent works in its own git worktree, preventing merge conflicts and allowing true parallel development across multiple features.

Rich Terminal UI​

Monitor all agents in real-time with a tiled terminal interface. Navigate between panes, restart stuck agents, and manage execution flow.

Quick Example​

tasks:
  - id: setup-auth
    title: Implement authentication service
    status: todo
    phase: 1
    repo: ./repos/backend
    worktree: feature/auth
    instructions: |
      Create a JWT-based authentication service with:
      - Login endpoint
      - Token refresh
      - Password hashing with bcrypt
    acceptance_criteria:
      - All auth endpoints return proper status codes
      - Tokens expire after 24 hours
      - Passwords are never stored in plain text

Next Steps​