LGTM alternativewhat to use now
LGTM.com, the automated code analysis service built by Semmle, was shut down in 2022 and its functionality moved into GitHub Code Scanning, powered by CodeQL. If you relied on LGTM, here is what the move covers, what it does not, and where an AI review layer fits.
Verdict
The short version
LGTM earned its reputation on depth: Semmle's CodeQL treated code as data and ran semantic queries across whole repositories, and the free tier for open source made it a fixture on countless README badges.
That chapter closed in 2022. GitHub had acquired Semmle, and LGTM.com's analysis lives on inside GitHub Code Scanning - CodeQL queries running on your repositories, with alerts surfaced on pull requests and in the Security tab.
Code Scanning is a worthy successor for what LGTM did: deterministic, query-based analysis with strong security coverage. What it does not do is reason about your specific change the way an AI reviewer does - the novel logic bug that no query pattern matches, the insecure refactor that typechecks cleanly.
The practical setup for most teams is both layers: keep Code Scanning for deterministic security gates, and add an AI reviewer for reasoning on the diff. CodeCritic is that layer - a GitHub Action, a paste-in-browser reviewer, and a REST API with a free tier.
Comparison
CodeCritic vs LGTM
LGTM.com shut down in 2022; the column describes the service as it was.
| Feature | CodeCritic | LGTM |
|---|---|---|
| Status | Active product with a free tier | Shut down in 2022 |
| Analysis model | AI reasoning on the diff, with context files and policies | CodeQL semantic queries over the repository |
| Where it runs | GitHub Action, web paste, REST API, VS Code and Cursor extensions | Hosted service at lgtm.com |
| Successor | CodeCritic | GitHub Code Scanning (CodeQL) |
| Findings style | Plain-language explanations with suggested fixes | Query alerts with severity and tracebacks |
| Free tier | Free tier on private repos, no credit card | Was free for open source projects |
Based on public docs and pricing pages as of October 2026 - verify against vendor pages before you buy.
What LGTM did well
Worth remembering, because the successor inherits most of it.
- Deep semantic analysis: CodeQL queries reasoned about data flow across the whole repository, not just patterns in single files.
- Free for open source, which made LGTM badges a quiet signal of project health.
- Deterministic results: the same query on the same code returned the same alerts, which auditors like.
- A clean alert workflow that fed straight into pull requests.
Where CodeCritic fits now
The layer Code Scanning leaves open.
- Reasoning about the change itself: logic bugs, broken invariants, and security issues that no query pattern anticipates.
- Review of code that is not in a connected repository - paste a diff or files in the browser, or call the API.
- Plain-language findings with suggested fixes, posted as PR comments or returned over the API.
- A free tier on private repositories, so you can evaluate on real code before committing.
When to stay with GitHub Code Scanning
Honest answer: for many teams, it is enough.
- Deterministic security gates are the requirement - CodeQL is auditable and repeatable in a way AI output is not.
- Your repos are public: Code Scanning is free and built in, with nothing to install or pay for.
- Your compliance process wants query-based evidence rather than model output.
What Code Scanning covers, and what it does not
Code Scanning runs CodeQL queries against your repository and surfaces alerts on pull requests and in the Security tab. For known vulnerability classes - injection, unsafe deserialization, path traversal - it is excellent, and it is already part of GitHub.
What it does not do is read your change the way a reviewer does. It will not notice that the new caching layer invalidates assumptions elsewhere, that the temporary flag bypass now guards auth, or that the error path leaks a token into logs. That is the reasoning layer, and it is where an AI reviewer earns its place.
FAQ
Common questions
Add the reasoning layer Code Scanning leaves open
Keep CodeQL for deterministic gates and run CodeCritic free on a private repo to see AI findings on your own pull requests.