Inline code review with MCP integration — add comments in the browser via SSH port forward, let Claude Code read and fix them via MCP tools.
Print a one-screen situational summary of the current code review state — open vs resolved comment counts, the files most-commented-on, recent reviewer activity, orphan tool calls, and threads where Claude has replied but the human hasn't acknowledged. Use when the user asks "what's the review status", "any open comments", "where am I in the review", "summary of the queue", "anything pending", or just wants to pick the work back up after a break. Read-only — does not edit files, reply, or resolve.
Read inline code review comments left by a human reviewer in the browser, then apply the requested fixes. Comments carry a scope (line / file / multi-file / view), one or more anchors (file + optional line range + blobSha + anchorText), and a viewContext saying which view the reviewer was in (tool-call / git-range / browse). Use when the user asks to "fix review comments", "apply review feedback", "read the review", or "address comments". Requires the code-review MCP server to be connected.
Read every open code review comment, group them by risk / file / theme, and present a phased fix plan for the user to approve before any edits run. Use when the review backlog is large (rough cutoff ≥ 5 open comments), when the user says "triage the review", "plan the fixes", "what's the plan for the review backlog", "group these comments", or before invoking review-workflow on a big batch. Requires the code-review MCP server to be connected. Read-only — does not edit files or call mark_resolved.
Admin access level
Server config contains admin-level keywords
Modifies files
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Hook triggers on file write and edit operations
Hook triggers on file write and edit operations
Tool-call review for Claude Code. Every Edit / Write / MultiEdit / NotebookEdit Claude Code runs is captured as a before→after snapshot, rendered in a browser UI as a timeline of cards. Leave inline comments on specific lines of either side, then let Claude Code read them via MCP and apply fixes.
Beyond the tool-call timeline, you can also create ad-hoc git diff review sessions (compare any two refs — branch / HEAD / SHA / INDEX / WORKTREE) and browse sessions (annotate any file in the worktree, no diff). Comments made in each mode are kept independent of one another.
Comments support four scopes: line (a specific line range), file (a whole file), multi-file (cross-cutting change across N files), and view (review-level / architectural note).
ssh -L 8080:localhost:8080 your-server
npx code-review-annotator --port 8080
One HTTP server can serve multiple projects / worktrees. The plugin's MCP server self-registers into a central registry; the browser UI lists every registered project in a dropdown.
When Claude Code is making changes on a remote machine, you want to see each edit it made individually — not squashed into a single git diff — so you can comment on decisions at the moment they happen. This tool closes that loop:
Edit / Write / MultiEdit / NotebookEdit — Pre/PostToolUse hooks capture the before and after snapshot"fix review comments" — it reads comments via MCP, applies fixes, and replies on each threadThere is no base branch, no merge-base. Each tool call is a standalone, frozen snapshot.
The plugin ships .mcp.json and hooks/hooks.json, so install via Claude Code's plugin commands:
/plugin marketplace add /path/to/code-review-annotator
/plugin install code-review-annotator@code-review-annotator-local
The plugin bundles:
claude mcp add neededEdit / Write / MultiEdit / NotebookEdit Claude Code runs# HTTP server (multi-project — projects register via the plugin's MCP)
npx code-review-annotator --port 8080
# Or pre-seed the registry with a specific project at startup
npx code-review-annotator --dir /path/to/project --port 8080
# MCP server (for projects that aren't using the plugin)
claude mcp add code-review -- npx code-review-annotator --mcp
Note: standalone MCP mode alone does not capture tool calls — you also need the plugin's Pre/PostToolUse hooks (or equivalent hooks wired manually into .claude/settings.json) for snapshots to appear in the UI.
npx code-review-annotator --port 8080
| Flag | Default | Description |
|---|---|---|
--port | 8080 | HTTP server port |
--dir | — | Optional in HTTP mode: if set, auto-registers this project at startup. |
--mcp | — | Run as MCP stdio server instead |
Projects register themselves into $XDG_CONFIG_HOME/code-review-annotator/projects.json (default ~/.config/...). The HTTP server also writes its current port into every registered project on startup, so the plugin's hook script knows where to POST snapshots. Stale entries (dir no longer exists) are filtered out of the UI automatically. The registry is safe to hand-edit.
ssh -L 8080:localhost:8080 your-server
open http://localhost:8080
The header has a three-way mode switcher:
Edit / Write / MultiEdit / NotebookEdit is a card. Click one to open its before→after diff. Newest at top. Status badges: pending / orphan.HEAD / INDEX (staged) / WORKTREE (current files incl. untracked), or paste a SHA. A warning appears if the two refs have no ancestor relationship — that case is outside this tool's design but still renders A..B. Each session is saved; switch between sessions via the sidebar dropdown..gitignore). No diff; pick a file, leave comments on any line.npx claudepluginhub jason-hchsieh/code-review-annotator --plugin code-review-annotatorDeveloper toolkit for the harness project. Currently provides skill-toolkit for creating, auditing, and improving Claude Agent Skills.
FEP/World-Models scaffolding for stable LLM coding agents: predict-before-act, inspectable belief state, and surprise-triggered reflection.
Send push notifications from Claude Code sessions via ntfy
Terminal-based code review companion for Claude Code. Review diffs, leave structured feedback, and submit reviews — all from a TUI running alongside Claude.
Native diff review window (Monaco + Glimpse) for AI coding agents. Browse changes since the merge-base with your base branch (or last commit / uncommitted), leave inline + file-level + overall comments on the slop, and have the agent address each item.
pair-review app integration — Open PRs and local changes in the pair-review web UI, run server-side AI analysis, and address review feedback. Requires the pair-review MCP server.
Automatic plan review with revdiff
Git diff review integration for Claude Code with VSCode extension. Provides inline diff view, browser diff editor, and auto-open files.
Cross-agent review workflow: Claude implements, Codex reviews