Skip to main content

Frequently Asked Questions

General

How are Semgrep and its rules licensed?

The Semgrep command line tool is open-source, licensed under LGPL 2.1 — you can use it at work, on private and proprietary code, no problem!

The Semgrep Registry contains rules from different contributors. Most rules, including all community-contributed rules, are under the MIT license. Semgrep Registry rules written by r2c are licensed under an LGPL 2.1 and Common Clause license.

The source for many r2c-written Registry rules is available at returntocorp/semgrep-rules. Those rules licensed under the LGPL 2.1 with Common Clause license cannot be resold without r2c’s permission. Since r2c offers a paid, hosted application (Semgrep App), it’s important to have this restriction so other companies, like major cloud providers, cannot resell r2c rules as a competing service.

Is it ok to run Semgrep or the r2c rules on my work projects?

Yes! Semgrep is safe to run on your private code. The Semgrep Registry license’s commercial restrictions only come into effect if you are selling the rules provided in the semgrep-rules repository. If that’s the case, you’ll need to talk with r2c first to get permission.

I’m a security professional and want to use the rules from the semgrep-rules repository with my clients as part of my paid services. Is that ok?

If you have a typical consulting service and running semgrep repository or semgrep-rules repository is part of your assessments, that’s great and of course feel free to refer your clients to the Semgrep App. But if your entire service is about scanning code and you want to charge for running the semgrep-rules repository that r2c and the community created and maintains, you need to discuss this with r2c.

Contacting r2c support

All users can contact r2c support. Regardless if you are free tier or paid tier user, reach our support through the r2c Community Slack. Semgrep Team tier customers receive 8*5 email/phone/Slack support with committed SLAs. See Support for more details.

Embedding the Playground in my website or blog post

Embed a special version of Semgrep Playground with an iframe. The source is https://semgrep.dev/embed/editor?snippet=<snippet-id> where the snippet-id is either the short identifier generated when you share a Playground link (this usually looks like DzKv) or the named identifier from a saved rule (this usually looks like username:rule-name).

<iframe title="Semgrep example no prints" src="https://semgrep.dev/embed/editor?snippet=DzKv" width="100%" height="432" frameborder="0"></iframe>


<iframe title="Semgrep example no prints" src="https://semgrep.dev/embed/editor?snippet=ievans:print-to-logger" width="100%" height="432" frameborder="0"></iframe>

Comparisons

How is Semgrep different from $OTHER_TOOL or $GENERIC_SAST?

Semgrep is an open-source tool with a simple syntax for writing rules: if you can write code, you can write a Semgrep rule — no program analysis Ph.D. required!

To our knowledge, the only other tool with the explicit goal of allowing custom rules is GitHub’s proprietary tool, CodeQL. CodeQL has a domain-specific language that is extremely powerful but is designed for those with significant program analysis expertise, whereas Semgrep is designed for the security engineer or developer who wants to automate code review. Our goal is to make writing a Semgrep rule as easy as copying the code you want to find—and letting the Semgrep engine make the rule and autofix high-quality enough to run in CI or your text editor or IDE.

Our hosted offering, Semgrep App, has a generous free tier (even for private repositories!). It offers a hosted CI integration with a quick setup so you can start running Semgrep right away. Semgrep's diff-awareness lets you scan new code and doesn’t force you to fix all the existing issues when you first start. For users running inside organizations with many repositories, the hosted offering also offers a policy and notification system that makes it easy to tune Semgrep so that it only reports issues or suggests fixes that get applied. Our goal is a 99% fix rate for what Semgrep reports.

Besides open-source and ease of writing new rules, what else is different about Semgrep?

1. Speedy & offline: Semgrep runs offline on every keystroke

If you are shipping code daily a code analysis tool that takes a week to run is not helpful. We think modern static analysis tools should run on every keystroke in the editor, without needing network access. Semgrep runs at approximately 20K-100K loc/sec per rule but our goal is to be even faster.

2. Semantic: Semgrep is smart

Semgrep automatically handles the nuance of “there’s more than one way to do it”: you write your query and all equivalent variations of that code are automatically matched.

As Semgrep evolves, queries similar to foo("password") become smarter. In the original version of Semgrep, this query would only match the code foo("password"). But a few months after release Semgrep would match const x = "password"; foo(x). Today Semgrep can do even more with intraprocedural dataflow analysis, and we’re working on adding more of these semantic features with every release.

3. Integrated: Semgrep understands git and other version-control systems

It’s easy to write a new Semgrep rule and have it only apply going forward. You can ignore findings of course, but we have built-in support for this with Semgrep CI and GitHub/GitLab/etc. integrations.

4. Portable: If you write a Semgrep rule, it runs anywhere

Many other tools require a buildable environment or can only be run in a VM. Semgrep runs “on the metal” and has minimal dependencies around a statically linked core; our parsers are declaratively-generated C libraries (we contribute to and use tree-sitter).

And many more: see the Semgrep philosophy for further reading.

Comparing Semgrep to linters

Linters use static analysis but typically have a narrower scope for analysis (most rules typically operate on a single line). Some linters also cover stylistic decisions (for example use of tabs versus spaces), but Semgrep doesn’t care about whitespace or formatting.

Semgrep’s registry includes rulesets inspired by the rules of many popular linters and checkers, including ESLint, RuboCop, Bandit, and FindSecBugs. But Semgrep also allows you to enable multiple rulesets at the same time without adding linter-specific artifacts or installation to your code repository.

Some popular linter tools may use tools like Semgrep as an internal engine, and we encourage this! For instance, the popular scanner NodeJSScan was re-written to use Semgrep as the core.

Lastly, while many linters are extensible, you need to learn specific abstract syntax tree (AST) based patterns for writing custom rules. Semgrep works across languages and you learn its syntax once; you don't have to mess with MemberExpressions, node visitors, and all that. Before Semgrep, many of us on the maintainer team were writing AST-based rules as well: one of us wrote an article comparing writing linter rules to Semgrep expressions.

Comparing Semgrep to CodeQL

Both Semgrep and CodeQL use static analysis to find bugs, but there are a few differences:

  • Semgrep operates directly on source code, whereas CodeQL requires a buildable environment
  • Semgrep is LGPL-2.1 and free to run anywhere; CodeQL is not open source and you must pay to run it on any non-open-source code
  • Semgrep supports autofixes; CodeQL does not.
  • Semgrep focuses on speed and ease of use. Because it doesn’t require a buildable environment, it doesn’t have some of the analysis features like interprocedural dataflow analysis that CodeQL does. (Semgrep does have limited intraproceedural dataflow).
  • Both have publicly available rules
  • Semgrep rules look like the source code you’re writing; CodeQL has a separate domain-specific-language for writing queries.
  • Semgrep has an online, hosted free plan; both have a hosted paid plan

See the Semgrep development philosophy for more about what makes Semgrep different.

Comparing Semgrep to SonarQube

Both Semgrep and SonarQube use static analysis to find bugs, but there are a few differences:

  • Extending Semgrep with custom rules is simple since Semgrep rules look like the source code you’re writing. Writing custom rules with SonarQube is restricted to a handful of languages and requires familiarity with Java and abstract syntax trees (ASTs).
  • Semgrep is LGPL-2.1, SonarQube offers an open-source version but it is missing features. For example, 12 of the supported languages are not available in the open-source offering, and more powerful dataflow features are only available in the paid versions.
  • Semgrep supports user-defined autofixes; SonarQube does not.
  • Semgrep focuses on speed and ease-of-use, making analysis possible at up to 20K-100K loc/sec per rule. SonarQube authors report approximately 0.4K loc/sec for rulesets in production.
  • Semgrep CI supports scanning only changed files (differential analysis), SonarQube does not
  • Both have publicly available rules
  • Semgrep has an online, hosted free plan; both have a hosted paid plan

See the Semgrep development philosophy for more about what makes Semgrep different.

Privacy and Security

Where do you store data?

r2c uses Amazon Web Services (US region) for storing customer data.

How is data secured, including data-at-rest and data-in-transit?

All customer data is located in AWS (US region). Amazon RDS encrypted database instances use industry-standard AES-256 encryption and TLS 1.2 or higher is used for all data-in-transit.

Is private source code shared with r2c?

No. Semgrep CI runs fully in your CI pipeline and your source-code never leaves your environment. Only meta-data related to Semgrep runs (see below) are sent to Semgrep's service.

What data is stored?

Semgrep may send data to Semgrep App in accordance with the metrics policy.

Configure Semgrep CI to explicitly send two types of data to Semgrep App. These types of data are scan data and findings data.

  • Scan data includes project name, CI environment, and scan meta-data.
  • Findings data are used to provide human-readable content for notifications and integrations, as well as tracking results as new, fixed, or duplicate.

For more information and a detailed description of each data field, refer to the relevant section in PRIVACY.md.

What network requests are made?

Semgrep CI makes network requests in accordance with the data storage mentioned above.

Semgrep makes the following network requests:

  • When running without --disable-version-check, Semgrep makes a network request to check for updates.
  • When providing a URL to --output, Semgrep performs an HTTP POST of the results to the specified URL.
  • When providing a registry ID like p/ci to --config, Semgrep requests the configuration from the Registry and may send metrics in accordance with the metrics policy.

Configuration

How do I configure Semgrep for different projects?

Semgrep App provides centralized policy management. See the Rule Board for more details.

What is a policy?

A policy is a simple collection of rules and a definition of what to do with rule results: fail the Semgrep CI run and/or send non-blocking notifications to third-party services like Slack. Please see the Rule Board for more details.

Monitoring

Do you have a visualization UI?

Dashboarding is available for Semgrep Team users. Semgrep also supports posting results through webhooks to any JSON endpoint, so you can easily integrate it with your favorite visualization tool.


Find what you needed in this doc? Join the Slack group to ask the maintainers and the community if you need help.