DAILY DROP · NOV 17, 2026

Today's drop: 1 new tool, 3 listing updates

The directory lists 150 hand-checked tools. New today: Open Code Review, Alibaba's free, open-source AI code review tool that reads your Git changes and leaves line-level comments, with built-in rules for security bugs such as cross-site scripting and SQL injection. Updated: Claude Code (v2.1.326), Qwen Code (v0.28.0) and CISO Assistant (v4.3.3). AI for IT & Cybersecurity grows to 65 tools; we retired one listing to stay at 150. Pricing only where the official page states it.

Previous drop: Nov 16, 2026 →

Get the weekly dropThe best new AI tools, one email a week. Free.

New in the directory 1

Browse all 150
Tool
Category
Price
Status
Link
Open Code Review (Alibaba)Free, open-source (Apache-2.0) AI code review command-line tool that started as Alibaba Group's internal code review assistant, which Alibaba says served tens of thousands of its developers over two years before it was opened up. Run ocr review in a repository and it reads your Git diff, groups related files, matches review rules to each file and has an agent read full files and search the codebase before writing line-level comments; ocr scan reviews whole files with no diff, for an audit of existing code. The built-in multi-language rule set covers bugs such as null-pointer errors and thread safety as well as security issues such as cross-site scripting (XSS) and SQL injection, and you can add your own rules by path. Two ways to run it: native mode uses any OpenAI- or Anthropic-compatible model endpoint you configure, and delegation mode lets the coding agent you already use (Claude Code, Codex, Cursor, Kimi Code or OpenCode) do the review under Open Code Review's file selection, rules and coverage checks, with no separate API key. Also runs in CI (GitHub Actions, GitLab CI, Gerrit), has a browser session viewer to mark findings fixed or ignored, and works on Windows, macOS and Linux (needs Git 2.41 or newer; install with npm install -g @alibaba-group/open-code-review). Alibaba's own benchmark (AACR-Bench, 200 real pull requests) reports higher precision than Claude Code on the same model at about a ninth of the tokens, with lower recall by design; those numbers are self-reported. About 53,400 GitHub stars (public since May 2026), about 181,700 npm downloads in the week of Nov 10-16, 2026, and frequent releases (v1.13.11 on Nov 17). An AI review is a second pair of eyes, not a replacement for a security scanner or a human reviewer.
Price Free and open source (Apache-2.0); you pay for your own model, or use your coding agent's model in delegation mode
FREE TIERchecked 11/17

Why this one

  • Open Code Review is the AI code reviewer with the most real-world use behind it that you can run for free. It started as Alibaba's internal review assistant, and since going public in May it has passed about 53,000 GitHub stars and about 180,000 npm downloads a week. Instead of asking a general agent to "review this", it decides which changed files must be reviewed, bundles related ones, matches review rules to each file, then lets an agent read the full code before it writes comments pinned to exact lines. The built-in rules include security checks such as cross-site scripting and SQL injection, and ocr scan reviews a whole folder with no diff, which is handy for auditing home-lab scripts or an old project. If you have no API key, delegation mode lets Claude Code, Codex, Cursor, Kimi Code or OpenCode do the review under its rules. It also runs in GitHub Actions and GitLab CI. Alibaba's own benchmark claims better precision than Claude Code at about a ninth of the tokens, with lower recall on purpose, so expect fewer false alarms but not every bug. Treat it as a second reviewer next to Semgrep or Snyk, not a replacement.

Listing updates

Checked on Nov 17, 2026 against official pages and release notes.

  • Claude Code: v2.1.326 (Nov 16) warns at startup when an --allowedTools, --disallowedTools or --tools entry does not look like a complete rule, warns and names the file when a permission rule you add cannot actually be saved, and stops describing Bash commands that delete or overwrite files (find -delete, cat a > b, sort -o) as a search or a read in the transcript. Code Mode now works with agent teams.
  • Qwen Code: v0.28.0 (Nov 17) runs headless sessions, the terminal UI and subagents on one shared agent loop by default, keeps the latest complete operation when a long session is compacted, shows per-turn hook statistics in the web shell and adds a DingTalk adapter. About 30,700 GitHub stars.
  • CISO Assistant: v4.3.3 (Nov 16) lets workflows fill in field defaults through inline event triggers and name the requester of a validation flow, keeps deprecated and draft items out of the landing-page briefing and fixes visual glitches in the library editor and metric charts.

Category changes

  • AI for IT & Cybersecurity grows from 64 to 65 tools with Open Code Review.
  • Audio drops from 6 to 5 tools: to stay at the 150-tool cap we retired Udio. Suno covers AI music generation for most of our readers, and our Udio listing never went beyond a one-line description. Its old page redirects to the category.

Checked, not listed

  • SkillEvaluator (NVIDIA, Apache-2.0, v0.9.0 on Nov 13): tests how agent skills and plugins change agent behavior; trending this week, but it is a quality framework for skill authors and SkillSpector already covers security checks of skills.
  • Laya Guard 0.8.0 (MIT, Nov 16) and Coral (withcoral, Apache-2.0): new allow/review/block gateways for agent tool calls; watching traction next to safeagents and claude-code-hooks. Vaikora Guard MCP: still watching.
  • Routine: SkillSpector 2.13.6 (false-positive fix for JSON metadata); Codex 0.171 alphas only; n8n 2.49.0 and 3.3.0 pre-releases. Ollama v0.40.17 and Wazuh 4.14.11 are still release candidates.