LoopForge: Orchestration of 10 Coding Agents with Human Approval Gates

A self-hosted workspace that manages the tasks performed by coding agent CLIs as team-level workflows. When requirements are provided via chat, it creates a Docker sandbox to check out the repository, delegates project roles to sub-agents, and automatically generates PRs. Repetitive tasks are saved as Standard Operating Procedures (SOPs) so that subsequent requirements are processed with the same steps and approval procedures.

It supports 10 coding agent CLIs, including CodeBuddy, Codex, Claude Code, and Cursor CLI, as executors. Independent prompts and models can be assigned to each stage, such as planning, implementation, review, and testing, ensuring clear role separation. All changes are linked to requirements and reflected in delivery and cost reporting.

It provides approval gates allowing human intervention during the automation process. You can check node status during execution, restart failed stages, or roll back to specific stages to continue work. Released under the MIT license, it can be deployed on your own infrastructure and supports various Git hosting services such as GitHub and Gitea.

It is suitable for development teams that require consistent quality control and audit tracking. A demo project is provided to allow immediate testing of script execution and approval flows without API keys. It is highly valuable in environments where you want to build a complex multi-agent collaboration structure while retaining final decision-making authority with developers.