Open source · AGPL-3.0

Your AI Development Team

Nine specialised AI agents covering the full SDLC — from ticket to production. Best practices built in. Every critical decision stays with you.

View on GitHub How it works ↓

pip install code-crew

Works with any LLM

NVIDIA Build

NIM endpoints

AWS Bedrock

Claude · Titan · Llama

OpenAI

GPT-4o · o1

Any OpenAI-compatible API

vLLM · Together · Groq

Langfuse

observability

The full SDLC, automated

From raw idea to production — agents do the heavy lifting at every stage, humans make the calls that matter.

Idea → Requirement

Turn any idea into a testable specification

→The crew reads your ticket, brief, or rough idea and gets to work immediately.
→The Product Owner and Scrum Master decompose it into user stories, acceptance criteria, and BDD scenarios — backed by your SOPs and domain glossary.
→Dependencies are mapped, ambiguities surfaced, and edge cases documented before a single line of design or code is written.
→The output is a structured requirement spec you can read, question, and refine.

Human gate

You review and approve the requirement spec. Nothing moves to architecture until you sign off.

Product OwnerScrum Master

The crew

Each agent has a distinct role, goal, and knowledge set. You lead; they execute.

👤you

Chief Architect

You — set priorities, approve designs, make the calls that matter. The crew works for you.

⚙️

Scrum Master

Sprint planning, ticket decomposition, flow coordination

🏛️

Architect

Design docs (ADDs/ADRs), component structure, dependency decisions

💻

Engineer

Full-stack implementation: backend, frontend, migrations, tests

🧪

QA Lead

BDD authoring, test strategy, acceptance criteria verification

📋

Product Owner

Requirement decomposition, BDD scenario review, DoD sign-off

🔒

Security Lead

OWASP review, OTM threat modeling, vulnerability triage

🚀

DevOps Lead

CI/CD coordination, infra changes, staging promotion

📦

Release Engineer

Release notes, launch go/no-go, smoke test verification

Commands

Every command works as a CLI call or a /slash command inside the interactive REPL.

explore [path]Scan project, detect stacks, build code index

design <KEY>Architect + security → ADD/ADR before any code

issue <KEY>Full SDLC: sprint planning → DoD → staging

sprint <name>Run all tickets in a sprint in parallel

threat [scope]OTM threat model + Threat Dragon JSON export

auditFull codebase audit: arch · security · compliance · domain

ask <agent> <q>Ask any agent a direct question

Quick start

From install to first implementation run in five steps.

Install

pip install code-crew

Configure LLM and issue tracker

cp .config.example.yaml ~/.code-crew/config.yaml
# set: llm provider, issue_tracker type, project key

Add a designs/ directory to your repo

# Option A — plain directory
mkdir designs

# Option B — git submodule (design docs in their own repo)
cd /path/to/your-repo
git submodule add git@github.com:your-org/designs.git designs

Initialise the project

code-crew init

Explore, design, implement

code-crew explore
code-crew design PROJ-123
code-crew issue PROJ-123

Open source — AGPL-3.0

Use, modify, and distribute freely. Any modified version you run as a network service must also be made available under AGPL-3.0.

Commercial license

Need to embed code-crew in a proprietary product or offer it as a hosted service without the copyleft obligation? Get in touch.