Plannotator
A local browser UI to annotate agent plans and review diffs, then send the feedback back to the agent
Install
curl -fsSL https://plannotator.ai/install.sh | bashInstalls the binary and configures detected agents. On Windows PowerShell: irm https://plannotator.ai/install.ps1 | iex.
This is third-party code. Review the repository files before installing.
What it does
Plannotator hooks into the agent through hooks and commands. When the agent writes a plan in plan mode, it opens in the browser, where you highlight passages, comment and send feedback to the agent. /plannotator-review shows uncommitted changes, GitHub PRs or GitLab MRs as a diff with line comments. /plannotator-annotate marks up Markdown, a folder, HTML or a URL, and /plannotator-last annotates the agent's latest reply. Data stays local, with encrypted share links for teammates.
Who it is for. For developers who review agent plans and code before letting changes through.
Good fit when
- You want to fix an agent plan with pinpoint comments instead of a long message
- You need to review the agent's diff or a PR and send notes back to the session
- You want a teammate to annotate a plan and pass their notes to the agent
Not a fit when
- You work on a remote server without a browser or port forwarding
- Regular review in the GitHub UI is enough
Example request
/plannotator-review https://github.com/org/repo/pull/42Limitations
Each UI surface checks GitHub for a new release on load, with no opt-out. URL annotation goes through Jina Reader by default, and AI features send context to your configured provider. PR review needs an authenticated GitHub or GitLab CLI. Features vary per agent; Droid has no plan interception.
How to disable. Run plannotator uninstall to remove hooks, skills and commands while keeping local plans and history. Add --purge to delete data too.
Security check
- Installs a binary and hooks into agent configs
- Reads project diffs and files and runs a local server
README in short
The README covers three flows: document and plan annotation, code review and HTML annotation. It lists annotate and review commands plus CLI commands for sessions and the decision archive. It details when data leaves the machine and how compressed and encrypted share links work. A single installer covers macOS, Linux and Windows with a per-agent follow-up step. Dual MIT or Apache-2.0 license.
FAQ
Which agents are supported?
Claude Code, Codex, Copilot CLI, Gemini CLI, OpenCode, Kiro, Droid, Amp and Pi. The installer detects them and configures hooks and commands for each.
Are my plans sent to a server?
No, everything is local by default. Content leaves only for link sharing, URL annotation and AI requests.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything
Up-to-date, version-specific library docs and code examples in your agent's context, via MCP or a CLI plus skill