AI Coding Agents · pcliangx

AppGenesisForge

AppGenesisForge is a Python-based scaffolding framework that orchestrates 19 AI agent roles (via Claude Code) through a 7-stage gated workflow to deliver features collaboratively. It enforces process discipline through mandatory skills, blocking hooks, and definition-of-done checklists rather than relying on agent autonomy alone.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/pcliangx/AppGenesisForge

737

GitHub stars

Forks

Python

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	pcliangx/AppGenesisForge
Owner	pcliangx
Primary language	Python
License	MIT — OSI-approved
Stars	737
Forks	16
Open issues	1
Latest release	v6.23.0 (2026-07-08)
Last updated	2026-07-08
Source	https://github.com/pcliangx/AppGenesisForge

What AppGenesisForge is

Built on Claude Code Agent Teams, AGF uses skill-based role assignment, git worktrees for isolation, TDD enforcement, 4-layer security hooks, and real-time kanban tracking. Supports multi-LLM backends (DeepSeek, Qwen, MiniMax, Doubao), full-stack web (React+FastAPI+Postgres), WeChat Mini Programs, and Apple native (Swift/SwiftUI for macOS/iOS).

Quickstart

Get the AppGenesisForge source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/pcliangx/AppGenesisForge.git2cd AppGenesisForge3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Multi-role AI-driven feature delivery with process governance

Teams needing parallel AI agent execution across development, review, testing, and deployment stages while enforcing TDD, code review discipline, and UAT sign-off. Ideal for organizations treating AI agents as a managed team rather than black-box automation.

Web full-stack + Apple native app development

Projects targeting React frontends with FastAPI/Postgres backends and/or native iOS/macOS apps using Swift. AGF provides role-specific templates, OpenAPI contract sync, and fastlane/notarytool integration for multi-platform release.

Regulated or quality-sensitive delivery with audit trails

Contexts requiring proof of TDD execution, SIT audit evidence, code review rationale, and UAT sign-off. The framework logs each stage gate completion and blocks progression without attestation.

Implementation considerations

Day-1 setup requires manual review of generated ADR-000 (system architecture) to prevent AI from inventing your tech stack. This is a critical gating step before team execution begins.
Requires Claude Code ≥ v2.1.154 and interaction via Claude's interface. No headless execution; terminal-driven commands (`/agf-team-start`, `/agf-board --watch`) integrate with Claude's agent orchestration.
Role cutdown during install (`setup/customize.sh --preset`) is optional but recommended for projects not targeting all tracks (e.g., omit WeChat Mini Program roles for web-only).
TDD enforcement is strict: commits without test-first evidence are rejected at code review. Teams must adopt red-green-refactor discipline or face workflow blockage.
Real-time kanban (`/agf-board --watch`) is HTML-based and relies on polling (~3 sec refresh); not suitable for sub-second monitoring or CI/CD pipeline integration.

When to avoid it — and what to weigh

Android native or Windows/Linux desktop apps — The framework explicitly excludes these platforms. No dedicated roles, templates, or release pipelines for Android or non-Apple desktop environments.
Single-engineer teams or rapid prototyping — AGF's 19-role orchestration, 7-stage gates, and enforced review/QA cycles add latency. For solo developers or MVP velocity, the overhead outweighs the governance benefits.
Real-time or ultra-low-latency requirements — Multi-agent orchestration and mandatory review gates introduce serialization points. Not suitable for systems requiring sub-second decision loops or continuous deployment without checkpoints.
Large language model training or fine-tuning — AGF is a delivery scaffold, not a model development platform. No support for training workflows, dataset management, or evaluation pipelines.

License & commercial use

MIT License — permissive open-source, explicitly allows commercial use, modification, and distribution with attribution.

MIT is an OSI-approved permissive license with no commercial restrictions. You may use AGF in proprietary products, closed-source forks, and commercial services without royalties. Retain the license header. No trademark or patent restrictions mentioned in the provided data.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

4-layer hook system: (1) dangerous command blocking (rm -rf, DROP TABLE, curl|sh), (2) secrets scanning for 11 vendors, (3) prompt-injection detection, (4) pre-commit diff scanning. No third-party pen test results or CVE history provided. Security relies on hook configuration and Claude's safety filters; actual efficacy requires review of `.claude/standards/security.md`. Teams must ensure credentials are not committed and are managed via Claude's safe context.

Alternatives to consider

Anthropic's built-in Claude Code features (no AGF scaffold)

Simpler, no framework overhead, but lacks role definitions, stage gates, TDD enforcement, and audit trails. Suitable for smaller projects or teams comfortable with agent autonomy.

Custom Agentic Workflow frameworks (LangGraph, CrewAI, AutoGPT clones)

More flexible and language-agnostic, but require custom implementation of role definitions, gates, and hooks. Steeper learning curve but no lock-in to Claude Code.

Traditional CI/CD + manual code review (non-AI-driven)

No LLM integration, slower development velocity, but proven governance. Better fit for teams wanting deterministic reproducibility over AI assistance.

Software development agency

Build on AppGenesisForge with DEV.co software developers

Start with the 5-minute quickstart, review the architecture ADR, and launch your first feature. Check FIRST_RUN.md for Day-1 checklist and troubleshooting.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

oh-my-taiyiforge SWE-AF mcp-context-forge

Related on DEV.co

Explore the category and the services that help you build with it.

AI Coding Agents AI Application Development Web App Development API Development Talk to DEV.co

AppGenesisForge FAQ

Do I have to use all 19 roles?

No. `setup/customize.sh --preset` allows role cutdown (e.g., omit miniapp roles for web-only projects). Roles map to delivery stages; smaller teams can collapse roles, but stage gates remain mandatory.

What if Claude Code becomes unavailable or changes?

Unknown. AGF is tightly integrated with Claude Code ≥v2.1.154 agent teams. No migration path to other platforms is documented. Dependency risk is present.

Can I use AGF with my existing repo?

Yes. `setup/install-to-existing.sh ~/path/to/repo` integrates AGF scaffolding into a live project. Existing code is not overwritten; ADRs and standards are merged. Day-1 `agf-init` validates and merges architecture decisions.

How does multi-LLM backend switching work?

Via `.claude/skills/agf-wiring-multi-llm-sdk/`. Credentials and API keys are managed in configuration. Switching requires updating skill parameters; no automatic fallback between vendors is mentioned.

Software development & web development with DEV.co

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If AppGenesisForge is part of your ai coding agents roadmap, our team can implement, customize, migrate, and maintain it.

Ready to govern your AI team?

Start with the 5-minute quickstart, review the architecture ADR, and launch your first feature. Check FIRST_RUN.md for Day-1 checklist and troubleshooting.

Talk to DEV.co Schedule a Consultation