Plannotator: review agent plans and diffs in your browser
Added NewCopyable prompt below
Published Last verified Released
Our verdict
Fixes the weakest part of agent work: reviewing a long plan or diff in a terminal and typing feedback from memory. Plannotator hooks into plan mode (or runs as a slash command), opens the plan, spec, HTML or diff in a local web UI, and turns your inline comments and markup into one message the agent receives as its next turn. Code review works on local changes or remote GitHub and GitLab PRs, and the newer Inbox gathers questions and reviews from several agents in one window without holding a session open. It has been public since December 2025, ships releases almost daily (v0.28.65 on Nov 15), and is GitHub Trending #1 this weekend at about 11,400 stars. Builders report real use: Chris Cook runs /plannotator-review before Claude's built-in /code-review on every merge, and @hanipcode reviews state machines with it alongside Herdr and Pi. Limits: the default install is curl | bash and wires hooks into each agent it finds (use --minimal for just the binary); every app view checks GitHub for updates with no opt-out; small share links put the plan in an unencrypted URL fragment (disable with PLANNOTATOR_SHARE=disabled); Workspaces is a separate hosted product with a waitlist. We have not run it ourselves.
- Who it's for
- Developers using Claude Code, Codex, Copilot CLI, Gemini CLI, OpenCode, Pi, Amp, Droid or Kiro who approve plans and review agent diffs every day
- Leverage
- Precise, line-level feedback on plans and diffs in a proper review UI, sent back to the agent in one click, so fewer bad plans get executed and fewer review rounds are lost to vague terminal replies
- Effort
- Low
- Date added
What it does
Plannotator (backnotprop/plannotator, Apache-2.0) installs a single plannotator binary and, per agent, the hooks, skills and slash commands that open its UI. Plan review intercepts plan mode in Claude Code, Copilot CLI and Pi, and uses Codex's experimental Stop hook. Annotate covers markdown, specs, HTML and URLs. Code review handles local Git, GitButler, Jujutsu and Perforce changes plus GitHub and GitLab PRs through your authenticated CLI, with comments and suggested code returned to the agent. Inbox (plannotator inbox) is a local window where agents leave questions, files and guided reviews; Claude Code, Pi and OpenCode connect through plugins and other agents over MCP (plannotator inbox mcp). The project says it collects no telemetry and keeps plans, annotations and history local by default; Ask A.I. features send context to the provider you configure.
How builders use it
- Install the binary and agent hooks with the official installer, or add the Claude Code plugin from the backnotprop/plannotator marketplace.
- Let your agent propose a plan in plan mode; Plannotator opens it in the browser. Comment and mark it up, then approve or send the feedback back.
- After the agent finishes, run /plannotator-review to review the diff (or a PR) with inline comments before your normal code review and merge.
- Running several agents? Open plannotator inbox and answer their questions and reviews in one place while they keep working.
What to expect
Plans get corrected before any code is written, and diff feedback reaches the agent as specific, line-anchored instructions instead of a paragraph typed from memory.
Copy this: Install + daily review loop
# Official source: github.com/backnotprop/plannotator (Apache-2.0)
# Full install (detects your agents and adds hooks, skills, commands):
curl -fsSL https://plannotator.ai/install.sh | bash
# Binary only, nothing wired into your agents:
curl -fsSL https://plannotator.ai/install.sh | bash -s -- --minimal
# Claude Code plugin route:
/plugin marketplace add backnotprop/plannotator
/plugin install plannotator@plannotator # then restart Claude Code
# Daily loop:
# 1. Plan mode -> Plannotator opens -> annotate -> approve or send feedback
# 2. /plannotator-review # review the diff in the browser
# 3. /code-review # then your usual review, then merge
# Optional:
plannotator inbox # one window for several agents
export PLANNOTATOR_SHARE=disabled # turn off share links
plannotator uninstallRelated discoveries
See the newest picks →- ToolCoding & Shipping Agents
improve: shadcn's skill where a strong model plans
An agent skill (MIT) from shadcn, the creator of shadcn/ui, that uses your most capable model to audit a codebase and write self-contained implementation plans, then hands those plans to cheaper models to carry out. It never edits your source code itself.
NewEffort: LowPrompt#Plugin / extension#Open source
- WorkflowCoding & Shipping AgentsHot
Run a codebase-wide audit with Claude Code dynamic workflows
Have Claude write a script that fans an audit out to many subagents, cross-checks findings, and can be saved and rerun.
Effort: MediumPrompt#Emerging
- ToolCoding & Shipping Agents
Cloudflare security-audit skill: agent bug hunts
Cloudflare's MIT skill that turns Claude Code, Codex or Cursor into a six-phase security auditor where a fresh agent tries to disprove every finding.
Effort: MediumPrompt#Plugin / extension#Open source
- ToolCoding & Shipping Agents
Elephant-Goldfish Model: specs reviewed by fresh agents
Google's Apache-2.0 plugin that makes your coding agent draft a design spec with you, then has fresh zero-context subagents attack it before any code is written.
Effort: MediumPrompt#Plugin / extension#Open source
Get the best new A.I. picks in your inbox
The best new agent workflows and tools we’ve vetted, with a clear verdict on each. Short, occasional, useful.