</>MCP Agents Market
Agent

ralphex

by umputun1.4kGoUpdated 2026-08-22

Extended Ralph loop for autonomous AI-driven plan execution

Claude CodeCodexGitHub Copilot CLIGeminiAntigravity

Ralphex is a standalone CLI tool that orchestrates autonomous execution of multi-step implementation plans using Claude Code or Codex. It solves the problem of manually supervising AI coding sessions by breaking plans into discrete tasks, running each in a fresh context to maintain quality, and automatically handling validation, commits, and multi-phase code reviews. Developers write a markdown plan with tasks and validation commands, start ralphex, and return to find their feature implemented, reviewed, and committed—or monitor progress in real-time via a web dashboard.

Key Features

Autonomous task execution with automatic retry, validation, and git commits after each task
Multi-phase code review pipeline with parallel agents (quality, testing, implementation, simplification, documentation)
External review integration with Codex, custom scripts, or local LLMs via configurable wrappers
Fresh Claude Code sessions per task to prevent context degradation and quality loss
Git worktree isolation for parallel plan execution without branch conflicts
Interactive plan creation via `--plan` flag with clarifying questions and editor feedback loop
Real-time web dashboard with SSE streaming, phase filtering, and multi-session monitoring
Docker isolation for safer autonomous execution with mounted credentials and project directory

Use Cases

  • 01Execute multi-task feature implementations overnight without supervising each Claude Code step
  • 02Maintain consistent code quality across long sessions by running each task in fresh context
  • 03Run automated multi-agent code reviews with configurable quality, testing, and security checks
  • 04Create structured implementation plans interactively with AI-guided context discovery
  • 05Monitor multiple concurrent plan executions across projects via web dashboard
  • 06Integrate alternative AI models (Codex, Copilot, Gemini) for task execution or review phases

Related Agents

View more

ralphex — FAQ

What is Ralphex and what does it do?+

Ralphex is a CLI tool that autonomously executes markdown implementation plans by orchestrating Claude Code or Codex. It breaks plans into tasks, runs each in a fresh AI session, handles validation and commits, and performs multi-phase code reviews without manual supervision.

How do I install Ralphex?+

Install via Homebrew (`brew install umputun/apps/ralphex`), from source (`go install github.com/umputun/ralphex/cmd/ralphex@latest`), or download binaries from GitHub releases. Docker wrapper script available for isolated execution.

Which AI clients does Ralphex work with?+

Ralphex orchestrates Claude Code and Codex CLIs by default. It also supports alternative providers via wrapper scripts including GitHub Copilot CLI, Gemini CLI, Antigravity (agy), OpenCode, and custom tools that produce stream-json output.

Does Ralphex require API keys or subscriptions?+

Ralphex requires either a Claude subscription (for Claude Code) or OpenAI/ChatGPT plan (for Codex mode). Note: starting June 2026, Anthropic's Agent SDK has separate credit pools from Claude Max subscriptions for unattended runs.

Is Ralphex free and open source?+

Yes, Ralphex is MIT licensed and free to use. However, the underlying AI services (Claude Code, Codex) require their respective paid subscriptions or API access.

What are the prerequisites for running Ralphex?+

You need a git repository, Claude Code CLI installed and authenticated (or Codex CLI for `--codex` mode), and optionally fzf for plan selection. Plans must be markdown files in `docs/plans/` with task sections and validation commands.

How do I install ralphex?+

Open the source repository on GitHub and follow its README. ralphex is a agent — MCP Agents Market links you directly to the official repo.

Is ralphex free?+

ralphex is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches