Overview
Re_gent and Deep Work Plan are two open-source tools designed to improve how developers work with AI coding agents, but they address different pain points. Re_gent focuses on version control for AI agent activity — tracking every change, prompt, and session so you can undo, blame, and replay agent work. Deep Work Plan, on the other hand, provides a structured methodology for long-horizon AI coding tasks, turning any repository into a harness with durable plans, acceptance criteria, and resumable state.
Feature Comparison
| Feature | Re_gent | Deep Work Plan |
|---|---|---|
| Core Purpose | Version control for AI agent actions | Spec-driven development harness |
| Approach | Git-like CLI (rgt log, rgt blame, rgt sessions) | Markdown-based methodology with skill pack |
| Agent Support | Claude Code, Codex, OpenCode (more planned) | Claude Code, Cursor, Codex, Copilot, Gemini, Windsurf, Cline, Antigravity |
| State Management | Content-addressed storage | Git-native, .dwp/ folder |
| Audit Trail | Per-line blame to exact prompt | Git-based, conformance check |
| Multi-Agent Work | Per-conversation branches | Agent-agnostic, resumable plans |
| License | Apache-2.0 | MIT |
Pricing
Both Re_gent and Deep Work Plan are completely free and open source. Re_gent is licensed under Apache-2.0 and promises to be "free forever." Deep Work Plan is MIT licensed. Neither has any paid tiers, usage limits, or hidden costs.
Pros and Cons
Re_gent Pros
- Provides a clear, git-like audit trail for AI agent actions
- Enables undo, blame, and replay of agent sessions
- Captures full conversation history even after agent compaction
- Free and open source with no usage limits
Re_gent Cons
- Limited agent support (only Claude Code, Codex, OpenCode currently)
- Requires CLI installation and learning new commands
- No built-in planning or task execution framework
Deep Work Plan Pros
- Agent-agnostic — works with many AI coding tools out of the box
- Provides a structured planning and execution framework with acceptance criteria
- Resumable state survives context resets; any agent can continue work
- Reasoning-based onboarding adapts to any repository's stack
Deep Work Plan Cons
- Requires initial setup and onboarding via init.md
- Methodology may have a learning curve for teams new to spec-driven development
- No built-in undo or blame for individual agent actions (relies on git)
Verdict
Choose Re_gent if you need a git-like version control system specifically for tracking and auditing AI agent actions, with undo and blame capabilities. Choose Deep Work Plan if you want a comprehensive, agent-agnostic framework for planning, executing, and resuming long-horizon AI coding tasks with structured acceptance criteria and validation.
Deep Work Plan
Re_gent