Skip to main content
Archived

This project has been archived by its maintainers, and is no longer receiving any updates.

Code Team Framework

This project implements the "Code Team Framework", a system that orchestrates a team of specialized AI agents to automate software development tasks. It operates on a "human-on-the-loop" principle, where the system handles the cycles of coding and verification, while strategic decisions are left to the user.

The framework is built to be stateless and resilient, deriving its state from the filesystem (Git status, plan files, logs) to allow for easy stopping and resuming of tasks.

Features

  • State Machine Orchestration: A central orchestrator manages the entire workflow, from planning to coding, verification, and committing.
  • Specialized AI Agents: A roster of agents (Planner, Coder, Verifiers, etc.) each with a specific role, powered by the Claude Code SDK.
  • File-Based State: The system is stateless. Stop it at any time and it will recover its state from the repository's condition.
  • Human-on-the-Loop: Key decision points, like plan approval and accepting code changes, require user intervention.
  • Configurable Workflow: Define verification steps, agent settings, and LLM providers in a simple .codeteam/config.yml.
  • Extensible: Designed with SOLID principles, making it easy to add new agents or verification steps.
  • Rich Terminal UI: Agent output is displayed in styled panels for better readability. Note: Scrolling within the panels is not supported - use your terminal's scrollback feature to review previous output.

Prerequisites

  • Python 3.10+
  • Node.js
  • Claude Code CLI: npm install -g @anthropic-ai/claude-code
  • An Anthropic API key set as an environment variable: export ANTHROPIC_API_KEY="your-key-here"

Installation

Install from PyPI (Recommended)

pip install code-team-framework

Install from Source

  1. Clone the repository:

    git clone <repository_url>
    cd code-team-framework
    
  2. Create and activate a virtual environment:

    python3 -m venv .venv
    source .venv/bin/activate
    
  3. Install dependencies:

    pip install -e ".[dev]"
    

Usage

The framework is controlled via the codeteam command.

1. Planning a New Feature

Start by creating a plan. The Planner agent will collaborate with you to break down the request into a task list.

You can provide the initial request directly as an argument:

codeteam plan "Implement a user profile feature with bio and avatar."

Alternatively, you can run the command without an argument to be prompted for the request interactively:

codeteam plan

The agent will ask clarifying questions. When you're ready, type /save_plan. The plan will be saved to .codeteam/planning/{plan_id}/plan.yml.

Review the plan and accept it to begin the coding phase.

# The tool will prompt you for this after the plan is saved
/accept_plan

2. Executing the Plan

Once a plan is accepted, start the coding and verification loop.

codeteam code

The orchestrator will pick up the next pending task, generate a prompt, and invoke the Coder agent. After the Coder finishes, verification checks run automatically. You will be prompted to review the changes.

  • Type /accept_changes to commit the work and move to the next task.
  • Type /reject_changes [your feedback] to send it back to the Coder with your notes.

3. Resuming Work

If you stop the process, you can resume it at any time. The orchestrator will automatically determine the current state and pick up where it left off.

# If you were in the middle of coding:
codeteam code

# If you were in the middle of planning:
codeteam plan

Integration

The Code Team Framework can be easily integrated into any project:

1. Initialize the Framework

# Navigate to your project
cd /path/to/your/project

# Initialize the framework (creates .codeteam/ directory)
codeteam init

This creates a .codeteam/ directory with:

  • config.yml - Main configuration file
  • agent_instructions/ - Customizable agent instruction templates
  • planning/ - Planning documents (generated during use)
  • reports/ - Verification reports (temporary)

Additionally, it creates .claude/commands/ with slash commands for manual workflows.

2. Start Using the Framework

# Start planning and coding
codeteam plan "Implement new feature"
codeteam code

3. Customization

After initialization, you can customize:

  • Configuration: Edit .codeteam/config.yml to adjust paths, verification commands, and agent settings
  • Agent Instructions: Modify templates in .codeteam/agent_instructions/ to customize how agents behave
  • Slash Commands: Use /codeteam-planner, /codeteam-coder, etc. in Claude Code for manual agent invocation

Configuration

The framework is configured via .codeteam/config.yml. You can set the LLM model, define custom verification commands, and configure agent behavior. Run codeteam init to create the default configuration file.

For comprehensive configuration options and examples, see the Configuration Guide.

Project Structure

Framework Source Code:

  • src/code_team/__main__.py: The command-line entry point (available as codeteam command after installation).
  • src/code_team/: The main application source code.
    • orchestrator/: The core state machine and orchestrator logic.
    • agents/: Implementations for all specialized AI agents.
    • models/: Pydantic models for configuration and plan files.
    • utils/: Helper modules for filesystem, Git, LLM interaction, and templating.
    • templates/: Default agent instruction templates (packaged with the framework).

When Used in Projects:

  • .codeteam/: Framework directory (created by codeteam init).
    • config.yml: Main configuration file.
    • agent_instructions/: Customizable agent instruction templates.
    • planning/: Generated plans and related documents.
    • reports/: Verification reports (temporary files).

Metadata

Release files for code-team-framework 0.2.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for code-team-framework 0.2.0
File Size Uploaded Provenance Actions
code_team_framework-0.2.0.tar.gz 60.4 kB Without provenance Details for code_team_framework-0.2.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for code-team-framework 0.2.0
File Interpreter ABI Platform Actions
code_team_framework-0.2.0-py3-none-any.whl Python 3 none any Details for code_team_framework-0.2.0-py3-none-any.whl

Total release size: 118.1 kB

Release files / code_team_framework-0.2.0.tar.gz

Download URL code_team_framework-0.2.0.tar.gz
Size 60.4 kB
Tags Source
SHA-256 checksum
How to use checksums
8f65bdca53015bf88bd0fc7f37aba8fc5884a06d9b0e7e1874f1d83efb0058e2
BLAKE2b-256 checksum
How to use checksums
baf9dd2524d75f252f978579261d3586cd42f15174592f52b7a02423c42d7974
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.12.3

Release files / code_team_framework-0.2.0-py3-none-any.whl

Download URL code_team_framework-0.2.0-py3-none-any.whl
Size 57.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
b8d1a8bb55cf8ad04fe05fdf12da8a171b911885df14a81a32a4903dda543b3c
BLAKE2b-256 checksum
How to use checksums
b850fc81df54390962f00efe3cf0151b6b9bd1281bd357512bb32252dc7de8d1
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.12.3

Security for code-team-framework 0.2.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page