ThinkReview — AI code review in the pull request sidebar

On the record

What people say after they use it

Reviews from the Chrome Web Store, Product Hunt, and G2, quoted as written. Cards link back to the listing.

Chrome Web Store Sep 13, 2026

Installed this expecting the usual AI review gimmick. It's not. I'm a backend dev, mostly PHP and Laravel, and I live in GitLab merge requests. Every other AI review tool I've tried had the same problem: an afternoon of wiring up CI and tokens, and then a bot dumps twenty comments on the MR that someone still has to sort through. ThinkReview has no setup at all. Install, open an MR, the review is already in the sidebar. No webhooks, no bot account, no pipeline config, nothing to rotate. Then I ran it on a real MR, not a demo, and it actually impressed me. It flagged a query that would have hurt at scale, an edge case I hadn't handled, and a couple of naming choices that would have confused whoever opened that file next. It tells you why, not just what, so it's easy to act on a suggestion or ignore it. Best part: nothing gets posted to the MR unless I post it. The review stays in my sidebar. It helps me review instead of reviewing for me.

Chrome Web Store Jul 3, 2026

Great extension for PR analysis, especially with Azure DevOps and a self-hosted gateway setup. The tool is really useful during code review; you can customize it or choose the LLM model according to your preferences when you are using AI Foundry. Overall, it makes reviewing PRs faster and more focused.

Product Hunt Pixel perfection

Came in expecting yet another noisy review bot, but the sidebar-only approach actually respects the workflow. Liked that I could point it at a local Ollama model and skip sending code anywhere, that part feels ahead of its time.

G2 Aug 11, 2026

Fast, Actionable Code Review Feedback That Catches Bugs Early

4.5/5

I like how ThinkReview delivers fast, contextual, and actionable feedback on code changes. It helps me catch potential bugs, edge cases, and maintainability issues early, while also reducing the manual effort involved in code reviews. The feedback is practical and easy to understand, which makes it a valuable addition to my development workflow.

Chrome Web Store Sep 1, 2026

As someone less technical and much more focused on product design, having thinkreview in my toolkit is a must. Firstly, it helps keep discipline and the codebase clean. Secondly, it catches issues that could easily be missed. It gives me an extra layer of confidence before things move further into production. For me, it’s simply one of those tools you should have in every development workflow.

Product Hunt Bright idea

Love that the reviewer stays in control of what actually gets posted to the PR, it's a thoughtful answer to the usual AI review tool fatigue.

Chrome Web Store Sep 20, 2026

One of the best AI tools focused on code quality that I've tested recently. It works really well in the development ecosystem, saves precious time, and the integration is super simple. I'm still learning its features in a clean and simple interface.

Chrome Web Store Sep 7, 2026

Really useful AI code review tool. It helps me catch potential issues and improve code quality before merging. The feedback is clear and easy to understand.

Chrome Web Store Apr 17, 2026

very handy to be able to chat wit the PR , I like the agents and the full context for the whole repo not just the diff.I use it for azure and bitbucket

From the sidebar to the whole repository

Scroll through what ThinkReview does: the sidebar, your editor, your rules, your agents, and every platform you run.

01 / 06

Lives in the sidebar.

ThinkReview opens beside the diff on the pull request page, so the review sits next to the code it talks about. No extra tab and no checkout of another branch.

A dark-theme GitHub pull request with the ThinkReview panel docked on the right, showing the PR description and a high-severity issue

02 / 06

Reviews from inside your editor.

Paste a pull request URL into Cursor, Claude Code, or Copilot. ThinkReview MCP fetches the changes and hands your assistant a structured review, using the same credits as the extension.

Cursor
Claude Code
Claude Desktop
VS Code Copilot

Get the setup config

Cursor reporting that ThinkReview ran on pull request 56 and that two real issues are fixed and pushed, followed by the list of fixes

03 / 06

Your standards, on every review.

Pick the security and best-practice checks your team cares about once, in the portal. They apply to every review, from the sidebar and from MCP, in both review layouts.

Custom rules

04 / 06

Set up an agent in the portal.

Name it, describe what it checks, add a reference, and choose when it runs: always, or only when the project or stack matches. Generate its output sections, then deploy.

Review agents

The Edit Agent dialog in the ThinkReview portal with a name, description, reference URL, run conditions and a preview of the agent's output sections

05 / 06

Agents run beside the review.

Every agent gets its own tab in the sidebar, next to the main review. Security, performance, or your own conventions, each with findings you can act on. An agent run uses one review credit.

The ThinkReview panel with Review, Performance analysis and Senior Pen Tester tabs, and a critical issue listed

06 / 06

Cloud or on-prem, same sidebar.

GitHub, GitLab, Azure DevOps, and Bitbucket, including GitHub Enterprise Server, self-managed GitLab, Azure DevOps Server, and Bitbucket Data Center.

GitHub

Cloudgithub.com
On-premGitHub Enterprise Server

GitLab

CloudGitLab.com
On-premSelf-managed GitLab

Azure DevOps

CloudAzure DevOps Services
On-premAzure DevOps Server

Bitbucket

CloudBitbucket Cloud
On-premBitbucket Data Center

A copilot for the reviewer. Not a bot in the repository.

Review bots post their opinion into the pull request for everyone. ThinkReview gives it to one person: the engineer who is accountable for the approval.

Review botsCodeRabbit, Copilot and similar

Everyone sees all of it

The bot talks to the whole pull request

Summary, walkthrough, diagrams and inline comments land in the thread, useful or not.
The author answers or dismisses each one in public.
Some tools can approve on their own, with no engineer reading first.

ThinkReviewVisible to one person: you

One click, one comment

The reviewer reads first and posts only what matters

Findings stay in a private panel beside the diff until you post.
One click posts the ones you choose, under your name.
ThinkReview cannot approve, merge or block. The human stays in the loop.

Human in the loopNo clutter in the pull request.

Left: an illustration of a typical AI review bot, such as CodeRabbit or Copilot, filling a pull request. Right: ThinkReview posting one comment.

Review pull requests without leaving your editor

Give Cursor, Claude Code, or Copilot a pull request or merge request URL. ThinkReview fetches the changes and returns a structured review to your assistant.

1

Add the server

One URL, mcp.thinkreview.dev/v1. With OAuth there is no token in your config file.
2

Authenticate once

Your client opens the ThinkReview portal in the browser. Sign in and approve.
3

Ask for a review

Paste a PR or MR URL and have your assistant call review_url_code.
Add a git personal access token under Settings → Integrations in the portal first. It works with GitHub, GitLab, Bitbucket, and Azure DevOps, and uses the same review credits as the extension.

Cursor → Settings → MCP

{
  "mcpServers": {
    "thinkreview": {
      "url": "https://mcp.thinkreview.dev/v1"
    }
  }
}

Enable the server, then click Authenticate.

Terminal

claude mcp add --transport http thinkreview https://mcp.thinkreview.dev/v1

Then run /mcp in Claude Code and choose Authenticate.

Claude Desktop → Settings → Connectors

https://mcp.thinkreview.dev/v1

Choose Add custom connector, paste the URL, and approve access in the browser.

.vscode/mcp.json

{
  "servers": {
    "thinkreview": {
      "type": "http",
      "url": "https://mcp.thinkreview.dev/v1"
    }
  }
}

Authenticate from the server in the MCP servers view, then enable it in Copilot Chat.

Then ask

Call review_url_code for https://github.com/org/repo/pull/123

Summary

Suggestions

Security notes

Best practices

Follow-up questions

Any other MCP client that supports remote HTTP servers with OAuth works the same way. Read the MCP integration guide

What teams ask before they install

Does ThinkReview post anything on its own?

No. The review stays in your sidebar. A comment goes onto the pull request only when you click Post. Depending on how you connect, it appears under your name or as ThinkReview through the GitHub App.

Is our code stored?

No. ThinkReview cloud processes the diff in memory and discards it. It is not stored and it is not used for training. If the diff must not leave your network, route reviews through the Enterprise Gateway or run them locally with Ollama.

Do we need admin approval or a pipeline change?

No. ThinkReview is a browser extension. There is no org-admin approval, no webhook, and no CI job to set up.

Which platforms are supported?

GitHub, GitLab, Azure DevOps, and Bitbucket, including GitHub Enterprise Server, self-managed GitLab, Azure DevOps Server, and Bitbucket Data Center.

Which browsers does it run in?

Chrome, Firefox, Edge, Brave, and other Chromium browsers.

Which models does it use?

ThinkReview cloud lets you pick from the latest frontier models on your plan and swap between them. Teams can point the Enterprise Gateway at their own model, and anyone can run reviews locally with Ollama. See all supported models.

Is it open source?

The browser extension is open source under AGPL-3.0 and developed in public on GitHub.

Review code like a senior.
From day one.

What people say after they use it

Fast, Actionable Code Review Feedback That Catches Bugs Early

From the sidebar to the whole repository

Lives in the sidebar.

Reviews from inside your editor.

Your standards, on every review.

Set up an agent in the portal.

Agents run beside the review.

Cloud or on-prem, same sidebar.

A copilot for the reviewer. Not a bot in the repository.

The bot talks to the whole pull request

The reviewer reads first and posts only what matters

Review pull requests without leaving your editor

Add the server

Authenticate once

Ask for a review

Simple Pricing

Choose where the review runs.

Where the diff goes

What teams ask before they install

What's New in ThinkReview

Loading latest version...

Latest from Our Blog

A developer can be reviewing in under a minute. A team can keep the code in-network.