HN Showcase
Back to browse
GitHub Repository

Local-first agent plan annotator

70 starsTypeScript

Open-Plan-Annotator – Annotate Agent Plans Like a Google Doc 100% Local

by ndom91·Feb 28, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemNiche Gem

Plan review loop with markup feedback—but agent planning UI is already Claude's bottleneck, not yours.

Strengths
  • •Genuinely solves friction in agentic workflows: human-in-the-loop plan approval with structured feedback sent back to the agent.
  • •100% local execution—no data leaves your machine, meaningful for users working with sensitive code or offline.
  • •Simple annotation UX (strikethrough, replace, insert, comment) mirrors Google Docs familiarity without overengineering.
Weaknesses
  • •Narrow audience: only works with Claude Code and OpenCode; Cursor, GitHub Copilot, and other agents don't integrate.
  • •Solves a problem that may not exist at scale—most developers approve agent plans without markup or reject wholesale.
Target Audience

Developers using Claude Code or OpenCode who want human-in-the-loop control over agent planning before implementation.

Similar To

Claude Code built-in plan review · GitHub Copilot's request changes flow

Similar Projects

Developer Tools●●Solid

Skill to annotate any Markdown file for AI feedback

You can hit /plannotator-annotate or run a local CLI to open a Markdown file in an annotation UI, make deletions/replacements/comments (even attach images), and export feedback as structured markdown that an agent can act on. It's a focused, practical tool for teams that already use agent plan modes — neat UX touches (browser UI + blocking CLI flow) but it stops short of broader integrations like PRs or VCS hooks that would make it indispensable.

Niche GemShip It
ramoz
104mo ago