Overview
Plannotator is an open-source, local feedback tool for coding agents. It turns questions embedded in a plan or Markdown document into cards, allowing a reviewer to answer them together on their own time and return the full set of responses to the agent in one message.
The same visual interface supports plan and document annotation as well as code review. A temporary local server opens the session in a browser, while integrations connect Plannotator to terminal and desktop coding agents.
Question Workflows
Questions can be presented in three forms: a single-choice question, a multiple-choice question, or a free-text question. A block can include context, selectable choices, an optional recommendation, and an already-settled choice for a later review round.
Reviewers may select an offered answer, provide another response, add a note, or skip a question. Returned feedback identifies answered and unanswered items, preserving the context needed by the agent. Question cards work in plan review, Markdown annotation, and the workflow for annotating an agent’s last message.
How to Use
The Plannotator installer can add a skill for agents that read agent skills, including Claude Code and Codex. The skill teaches the agent the supported question syntax; alternatively, teams can place the documented question format in their own agent instructions.
A typical round follows a concise sequence:
- The agent writes decision questions into a plan or Markdown file.
- Plannotator opens the questions as cards in the browser.
- The reviewer answers the cards and sends the responses together.
- The agent receives structured answers, notes, and any unanswered items.
Plans, Documents, and Code Review
Beyond questions, Plannotator can annotate Markdown, HTML, plans, documents, folders, and supported local files. Reviewers can select text, request deletions, attach comments or images, edit Markdown, and send structured feedback.
For code, it can inspect local Git, GitButler, Jujutsu, and Perforce changes, as well as GitHub pull requests and GitLab merge requests. The review interface supports comments on exact lines or ranges and replacement suggestions; local review does not commit, merge, or deploy changes. Optional configured AI reviewers can also analyze a change, with their findings intended to be verified against the current code and revision.
Local Operation and Compatibility
Normal local sessions bind to the computer’s loopback address, and Plannotator stores plans, drafts, history, configuration, and session records in its data directory. The open-source edition does not send usage telemetry or analytics, although release checks and optional provider, URL, sharing, hosted, or AI features may contact external services.
Plannotator supports multiple coding-agent environments through host-specific hooks, plugins, commands, or skills, including Claude Code, Codex, OpenCode, Copilot CLI, Gemini CLI, Pi, Amp, Droid, and Kiro CLI. It is dual-licensed under MIT or Apache 2.0.