Panel rating · 6 judges · How to read the stars

Category median

Sovereignty: 5 of 6 dimensions proven

0–5 in half steps. 5 means the rubric's top anchor is met on the evidence.

by Semgrep, Inc. · semgrep.dev

Compare with Snyk → Report an error on this page Is this your product? →

Link to this page

Semgrep: AI-assisted analysis on whats-best.ai Semgrep: AI-assisted analysis on whats-best.ai (dark)

Add ?theme=dark to the image address for dark backgrounds. The badge shows the category and the month of the latest evaluation, and EU-Made where that is the product’s provenance. It never shows a score.

Semgrep at a glance

Vendor
Semgrep, Inc.
Headquarters
Incorporated in the United States source, captured 6 Nov 2026
Data hosting
Outside the EU by default; pages mention the United States source, captured 6 Nov 2026
Free plan
Yes, per the vendor’s own page source, captured 6 Nov 2026
Panel verdict
The AI panel rates Semgrep 3.5/5, the mean of 6 judges. Semgrep is a SAST, supply-chain and secrets platform, strongest on developer workflow, where five of six judges scored 8: findings arrive as PR and MR comments with in-thread triage, merge-blocking policies and an autofix that opens a draft pull request in 2 to 10 minutes. how scores are made
As of
Vendor pages captured 6 Nov 2026; evaluated by the panel 6 Nov 2026. all sources

Semgrep alternatives, European ones first →

Read this page as one judge. Each weighs the same scores by what they care about.

The panel's verdict

Semgrep is a SAST, supply-chain and secrets platform, strongest on developer workflow, where five of six judges scored 8: findings arrive as PR and MR comments with in-thread triage, merge-blocking policies and an autofix that opens a draft pull request in 2 to 10 minutes. Dependencies and code analysis follow at 7 to 8, resting on reachability analysis for eleven ecosystems written into CycloneDX 1.4-1.7 SBOMs and per-language cross-file taint analysis in a public rule language. Trust and compliance scores run 6 to 7, pairing a stated-period SOC 2 Type II report with terms that grant a perpetual, sublicensable licence to use Customer Data to improve the vendor's products. Sovereignty is the flat-4 weakness: Semgrep, Inc. is a Delaware corporation under California law, servers are stated to be in the United States on AWS, and we found no public information on EU hosting or an EU contracting entity. Pricing transparency scores run 5 to 6 on an Enterprise tier quoted only as volume pricing and an undefined contributor count. flagged splits: none; persona-weighted totals run 6.3 to 7.2.

Report an error

Speaks for it

  • Cross-file taint analysis is documented per language, with source-to-sink traces and a public rule language for writing custom rules
  • Reachability analysis is generally available for eleven ecosystems and written into CycloneDX 1.4-1.7 SBOMs with a per-vulnerability state and justification
  • Secrets detection covers 630+ credential types, with validity checks that run locally on customer infrastructure and pre-commit hooks
  • Findings arrive as PR and MR comments with in-thread triage commands, merge-blocking policies and autofix that opens a draft pull request in 2 to 10 minutes
  • CLI and customer-managed CI/CD scans run on customer-controlled infrastructure, so source code need not leave the customer's machines

Report an error

Held against it

  • Sovereignty scores a flat 4 — a Delaware contracting entity under California law, servers stated to be located in the United States, and we found no public information on EU hosting or an EU contracting entity
  • The Enterprise tier is volume pricing with no published figure, and how a contributor is counted is not explained, so an annual cost is not computable from the public pages
  • The terms grant a perpetual, sublicensable licence to use Customer Data to improve and develop the vendor's products, and we found no public information on a code retention period or a published SLA
  • We found no public information on SCIM provisioning, SARIF or CSV export of findings, or granular roles per team
  • The published accuracy figures — triaging 80% fewer false positives and up to 3.5x more true positives vs AI alone — are the vendor's own, with no independent benchmark on the pages read

Report an error

Best for

  • You need deep, customizable SAST — cross-file taint analysis with a documented rule language for your in-house frameworks and injection rules
  • Your developers fix findings in code review — PR and MR comments with triage commands, merge-blocking policies and autofix that opens a draft pull request in 2 to 10 minutes
  • Your supply-chain work needs focus — reachability analysis for eleven ecosystems and CycloneDX SBOMs recording whether the code is actually affected
  • You want scans on your own infrastructure — CLI and customer-managed CI/CD deployments run on customer-controlled infrastructure, with an open-source Community Edition under LGPL 2.1

Report an error

Avoid if

  • You require EU hosting, an EU contracting entity or an EU representative — the documented chain is US-based and we found no public information on EU options
  • You require a vendor that is itself ISO 27001 certified — the pages state plainly that Semgrep Inc. is not
  • Your legal review cannot accept a perpetual, sublicensable licence to use Customer Data to improve and develop the vendor's products
  • You must compute a firm annual cost before committing — the Enterprise tier is sold as volume pricing and we found no explanation of how a contributor is counted

Report an error

The scores

Panel median The median of the judges’ scores per criterion. Each label opens its reasoning below.

Code analysis (SAST & code quality)

Show reasoning
How this is scored

What the tool finds in the company's own code: static application security testing with data-flow and taint analysis across functions and files, the languages and frameworks covered and at what depth, the rule sets (OWASP Top 10, CWE, custom rules), bugs and code smells, duplication, cyclomatic and cognitive complexity, maintainability and technical-debt measures, test coverage import, infrastructure-as-code checks, and how false positives are measured and kept down.

0 — Does not analyse source code; a dependency, secret or runtime scanner only, or a different kind of product.

3 — Pattern-based rules or linters for a handful of languages; data-flow analysis, the rule list, quality metrics and false-positive handling are basic or not described.

5 — SAST for the common languages (Java, JavaScript/TypeScript, Python, C#, Go, PHP) with the rule list published and mapped to OWASP and CWE, bugs and code smells, duplication and complexity metrics, coverage import, and custom rules or rule configuration documented.

8 — Cross-file taint analysis with the supported frameworks listed per language, a rule language customers can write in, infrastructure-as-code checks, maintainability or code-health scores over time, false-positive rates or benchmarks published, and each finding explained with its data flow and a fix.

10 — Analysis a large polyglot codebase relies on: deep interprocedural analysis across many languages with the depth stated per language, independent benchmark results, quality and security trends per team and hotspot, and every rule documented with examples and its precision.

Report an error

The Lead Developer

Cross-file taint analysis is documented per language — Java, JavaScript, TypeScript, Python, Go and C# get cross-file dataflow while PHP gets cross-function, with Pro rule counts per language and Laravel among the named frameworks, and the taint-mode rule language is public so we can write our own injection rules. Terraform checks are generally available and findings carry a taint trace from source to sink plus autofix or a rule-defined fix. The published precision figures (80% fewer false positives triaged, 3.5x more true positives versus AI alone) are vendor claims, and we found no public information on duplication, complexity or coverage metrics. 14 15 6 1 2 18

Report an error

The Application Security Engineer

Cross-file taint analysis is documented per language with source-to-sink traces and the interfile engine behind a pro flag, coverage is stated per language with Pro rule counts and framework-specific control flow for Java, JavaScript and TypeScript, Terraform checks are generally available, and the rule language is deep enough to encode our in-house frameworks. What holds it below the top: the false-positive figures are the vendor's own claims ('80% fewer false positives', 'up to 3.5x more true positives at 19% lower cost per true positive versus AI alone') rather than independent benchmarks, and we found no public information on quality metrics such as complexity, duplication or technical debt. 1 2 5 6 14 15

Report an error

The Engineering Manager

Per-language depth is published: Java and Kotlin show cross-file dataflow analysis with 190+ and 60+ Pro rules, taint findings carry a source-to-sink trace, the taint-mode rule language is documented deeply enough for my teams to write their own rules, and Terraform and Dockerfile checks exist. What I found no public information on is the code-health side I need for a ten-year Java codebase — maintainability, complexity, duplication or technical-debt measures and per-team or hotspot trends — and the true-positive and false-positive figures quoted (3.5x more true positives, 80% fewer false positives triaged) are the vendor's own rather than independent benchmarks. 15 14 6 2 1

Report an error

The Platform Engineer

Static analysis depth is documented per language — 35+ languages, cross-file dataflow analysis for Java, Python, JavaScript, TypeScript, C#, Go and C/C++, framework-specific control-flow analysis, and Pro rule counts per language such as 710+ for Python and 250+ for JavaScript. Taint analysis is documented down to sources, sinks, propagators and sanitizers with interfile analysis, findings carry the taint trace from source to sink, rules are written in a documented rule language, and Terraform checks are generally available. The false-positive figures (80% fewer false positives, 3.5x more true positives versus AI alone) are the vendor's own claims rather than independent benchmarks, and we found no public information on quality metrics such as duplication or complexity, or on test coverage import. 15 14 5 6 1

Report an error

The Product Security Officer

Cross-file taint analysis is documented language by language with framework-specific control flow, per-language rule counts, a documented rule language customers write in, and Terraform and Dockerfile coverage, with each taint finding explained by its trace — the depth an auditor wants to see. What holds it back: we found no public information on maintainability, complexity or coverage metrics, and the false-positive figures (80% fewer triaged, 3.5x true positives versus AI alone) are vendor claims rather than independent benchmarks. 2 6 14 15

Report an error

The Skeptic

The depth is real where it counts: cross-file taint analysis is documented per language with a rule language customers write in (pattern-sources, pattern-sinks, propagators, sanitizers), taint traces showing the flow from source to sink, framework-specific control flow for Java, Python, JavaScript and TypeScript, per-language rule counts (Python 710+, JavaScript 250+), and Terraform for infrastructure-as-code. It stays below the top levels because we found no public information on complexity, duplication, coverage import or any code-health trend over time, and the accuracy figures — "triage 80% fewer false positives" and "3.5x more true positives" — are the vendor's own numbers with no independent benchmark on the pages we read. 1 2 5 6 14 15

Report an error

Dependencies (SCA), licences & secrets

Show reasoning
How this is scored

What comes from outside or should never be committed: software composition analysis of direct and transitive dependencies with the ecosystems and lockfiles supported, the vulnerability database and how quickly it is updated, reachability analysis, licence detection and policies, SBOM export (CycloneDX, SPDX), malicious and typosquatted package detection, end-of-life components, automatic upgrade pull requests, container image scanning, and secret detection in code and git history with the number of detectors, validity checks and revocation guidance.

0 — No dependency or secret scanning of any kind.

3 — Known-vulnerability alerts for dependencies in one or two ecosystems or a basic secret pattern check; transitive dependencies, licences, SBOMs and git history are basic or not described.

5 — SCA for the main ecosystems including transitive dependencies, licence detection with policies, SBOM export in a standard format, upgrade suggestions or pull requests, and secret detection in commits with the detector list described.

8 — Reachability analysis that shows whether vulnerable code is called, malicious-package detection, the vulnerability database and its sources documented, end-of-life components flagged, secrets checked for validity and scanned across the full git history, and push protection that stops a secret before it lands.

10 — Supply-chain coverage a regulated software maker can file: every ecosystem the company builds with, reachability for each, SBOM and VEX output for customers, licence obligations explained per licence, secrets found outside git (tickets, chat, containers) with automated revocation, and the database's coverage and latency evidenced.

Report an error

The Lead Developer

Reachability analysis is generally available for eleven languages including PHP and JavaScript/TypeScript, with malicious-dependency detection and EPSS alongside CycloneDX SBOMs in versions 1.4 to 1.7 that carry CWEs, advisory links and a reachability state with justification. Licence policies draw on deps.dev plus Packagist for PHP and Rubygems for Ruby, block or comment in the pull request with per-dependency exemptions, and dependency upgrade pull requests are automated. Secrets cover 630+ credential types with local validation (the secret never goes to Semgrep's servers), pre-commit hooks and history scanning in beta; we found no public information on end-of-life component flagging. 6 15 16 17 3 8 18

Report an error

The Application Security Engineer

Reachability analysis is generally available for eleven ecosystems and written into the SBOM as a state and justification — exactly what keeps a backlog holding only code that is actually called — with malicious dependency detection, EPSS, licence policies per category with allow/block decisions, and autofix pull requests that upgrade vulnerable dependencies. Secrets detection covers 630+ credential types with semantic and entropy analysis, validity checks against services like AWS, Slack and GitHub performed locally so the secret never leaves my infrastructure, and pre-commit hooks that stop a leak before it lands; historical scanning of the git history is documented but marked beta, and we found no public information on end-of-life component flagging or container image scanning. 3 6 8 15 16 18

Report an error

The Engineering Manager

Reachability analysis is generally available for eleven ecosystems including Java, Kotlin, JavaScript and TypeScript, malicious-dependency detection and EPSS are listed, CycloneDX 1.4 to 1.7 SBOMs carry CWE numbers and reachability justifications, licence policies can block a pull request using deps.dev data, and secrets cover 630+ credential types with validity checked locally and pre-commit hooks. End-of-life components, container image scanning, SPDX export and the vulnerability database's own documentation I found no public information on, and scanning secrets across the full git history is marked Beta. 6 8 16 17 3

Report an error

The Platform Engineer

Supply-chain analysis covers 14 languages with reachability analysis generally available for eleven of them including Java, Go, Python and C#, malicious dependency detection, EPSS scores, licence allow/block policies with exemption lists and named licence sources (deps.dev, Packagist, Rubygems), CycloneDX 1.4–1.7 SBOMs carrying CWE numbers and reachability state fields, and an autofix that opens upgrade pull requests. Secret detection covers 630+ credential types with semantic and entropy analysis, validity checks that run locally within our own infrastructure, a pre-commit hook to stop secrets landing, and historical scanning in beta. We found no public information on container image scanning or end-of-life component flagging. 15 16 17 18 3

Report an error

The Product Security Officer

Reachability analysis with the affected languages listed, malicious-dependency detection, SBOM generation in CycloneDX 1.4 through 1.7 with per-vulnerability reachability state and justification, licence compliance that can block a pull request with exemptions, secrets validated against the issuing service locally, and pre-commit protection — close to what a regulated software maker files. Historical scanning of the git history is labelled beta and we found no public information on end-of-life component flagging. 3 6 8 16 17

Report an error

The Skeptic

Reachability analysis is generally available for eleven ecosystems and the documentation states outright that Dart and Elixir have none — a stated limitation I respect — alongside malicious dependency detection, EPSS scoring, CycloneDX SBOMs in versions 1.4 through 1.7 with per-vulnerability reachability state and justification, and licence policies that block or allow with exemptions and named licence sources. Secrets detection covers 630+ credential types with validity checks that run locally on the customer's own infrastructure, pre-commit hooks, and historical scanning explicitly marked Beta; we found no public information on end-of-life component flagging or container image scanning. 3 6 8 15 16 17

Report an error

Developer workflow, quality gates & fixes

Show reasoning
How this is scored

Whether developers see and act on the findings where they work: pull and merge request comments and status checks, quality gates that block a merge on new issues, a focus on new code against a baseline, IDE plugins with the same rules as the server, CLI and CI integrations (GitHub Actions, GitLab CI, Jenkins, Azure Pipelines, Bitbucket Pipelines), pre-commit hooks, triage, suppression and ignore rules with reasons, deduplication, fix suggestions and AI autofix with the model behind it, and scan times.

0 — Findings only in a separate report; no pull request, IDE or CI integration.

3 — A CI step or a dashboard of findings; pull request comments, gates, IDE plugins and triage are basic or not described.

5 — Pull or merge request comments and status checks on the common platforms, a configurable quality gate that can fail a build, a new-code baseline, IDE plugins for the common editors, CI integrations documented, and findings that can be ignored or marked as false positives.

8 — IDE results that match the server rules, gates per branch or project with thresholds per severity, pre-commit or push-time checks, deduplicated findings across tools, fix suggestions or AI autofix as a reviewable change with the model provider stated, and typical scan times documented.

10 — A workflow that keeps noise low on a large team: findings routed to the owning team, every suppression audited with a reason and expiry, autofix with its acceptance rates published, and the gate's behaviour on every supported platform documented with its limits.

Report an error

The Lead Developer

Findings arrive as pull or merge request comments with triage commands, ignore reasons and cross-branch triage, policies can block a merge with an auditable developer override, and diff-aware scans focus on new code — exactly the loop I want before review. IDE plugins for VS Code and JetBrains share the same Pro and OSS engines as the rest of the products, pre-commit hooks stop secrets before they land, and Autofix opens a draft pull request within 2 to 10 minutes with Amazon Bedrock and Anthropic Claude named as the model. We found no public information on deduplication across tools or full scan-time documentation beyond the secrets page. 5 19 18 8 3 21 22

Report an error

The Application Security Engineer

Developers get PR and MR comments on GitHub, GitLab, Bitbucket and Azure with reply-to-triage commands and named reasons, diff-aware scans against a baseline, remediation policies that can block a merge and are overridable only with permissions, noise control so only rules meeting severity criteria comment, IDE plugins for VS Code and JetBrains, and AI autofix opening a reviewable draft PR in 2 to 10 minutes with the model provider named as Amazon Bedrock and Anthropic Claude. Held below the top because we found no public information on whether the IDE plugins run the same rules as the server, whether suppressions carry an expiry, or deduplication of findings across tools. 5 8 18 19 21 22

Report an error

The Engineering Manager

This is where it fits my teams: pull and merge request comments on GitHub, GitLab, Bitbucket and Azure, diff-aware scanning, remediation policies that can block a merge when a rule fires (overridable with the right permissions), triage from PR comments and a bulk triage API with reasons, and Jira ticketing routing. Autofix opens a draft pull request in 2 to 10 minutes with the model behind it named — OpenAI by default, Amazon Bedrock with Anthropic Claude, or a custom provider — and Guardian scans AI-written code, which is exactly the number I owe the board. Scan-time figures are thin, and suppression expiry and autofix acceptance rates I found no public information on. 5 19 18 21 22 1

Report an error

The Platform Engineer

Findings reach developers where they work: pull and merge request comments across GitHub, GitLab, Bitbucket and Azure with Jenkins, CircleCI and Buildkite in the CI list, triage directly from review comments with reasons such as false positive and acceptable risk, a nosemgrep comment, and diff-aware scans against a baseline. Remediation policies can block a merge when conditions on severity and confidence are met, with permission-gated overrides, and pre-commit hooks stop secrets before commit; Autofix opens a draft pull request with the fix in 2 to 10 minutes, with Amazon Bedrock and Anthropic Claude named as providers and OpenAI the default for the multimodal features. We found no public information on autofix acceptance rates or on routing findings to owning teams. 5 19 18 3 22

Report an error

The Product Security Officer

Findings arrive as pull and merge request comments with triage commands and documented ignore reasons, remediation policies can block a merge with a documented developer override, pre-commit hooks stop secrets before they land, and AI autofix opens a reviewable draft pull request in 2 to 10 minutes with the model providers named (OpenAI, AWS Bedrock with Anthropic Claude). Typical scan times are documented only for secrets and autofix, and we found no public information on deduplication of findings across tools. 3 5 18 19 22

Report an error

The Skeptic

Findings reach developers where they work: PR and MR comments across GitHub, GitLab, Bitbucket, Jenkins, CircleCI, Azure and Buildkite, triage from the comment thread with a fixed set of documented reasons, remediation policies that can block a merge (the licence check exits with code 1) and control which rules may comment, pre-commit hooks, and an AI autofix that opens a draft pull request in 2 to 10 minutes with its model provider named — Amazon Bedrock, including Anthropic Claude models. What keeps it off the top level: no published autofix acceptance rates, and suppression carries reasons but we found no public information on expiry or audit of suppressions. 3 5 17 18 19 21 22

Report an error

Platforms, deployment & administration

Show reasoning
How this is scored

How the tool fits the company: the code hosts supported (GitHub, GitLab, Bitbucket, Azure DevOps) cloud and self-managed, a self-hosted or on-premise edition or a scanner that runs inside the customer's network, a documented REST API and webhooks, issue tracker integrations (Jira), single sign-on (SAML or OIDC) and SCIM, roles and permissions per project and team, portfolio reports across repositories, audit logs, export of findings (SARIF, CSV), and partner or multi-organisation accounts.

0 — One repository at a time with no integrations and no administration.

3 — A web dashboard and one code host; the API, roles, SSO and self-hosted options are basic or not described.

5 — GitHub, GitLab and Bitbucket or Azure DevOps supported, a documented REST API, Jira integration, roles per organisation and project, single sign-on, reports across repositories, and export of findings in a standard format.

8 — Self-managed code hosts supported, a self-hosted edition or an in-network scanner documented, SCIM provisioning, audit logs, granular roles per team, portfolio dashboards with trends, SARIF import and export, and webhooks for findings.

10 — Administration that scales across many teams and business units: every setting reachable by API with parity stated, an air-gapped deployment documented, SSO enforced with SCIM, audit logs exportable to a SIEM, and multi-organisation or partner management.

Report an error

The Lead Developer

All four major hosts plus Jenkins, CircleCI and Buildkite are covered, self-managed Bitbucket Data Center and on-prem source code management are supported, the REST API is documented in OpenAPI with stated maturity levels for v1 and v2, and Jira, Slack, email and finding webhooks are there. SSO via OIDC and SAML on Teams and Enterprise, audit logging, unlimited repositories on Enterprise and MCP integrations for AI tools round it out, with the CLI able to run scans on customer-controlled infrastructure. We found no public information on SCIM provisioning, SARIF export or granular per-team roles, which keeps it below the top band. 5 7 20 8 21 9 1

Report an error

The Application Security Engineer

The platform documents PR checks across GitHub, GitLab, Bitbucket and Azure, self-managed hosts including Bitbucket Data Center and on-prem source code management at the enterprise tier, an OpenAPI-documented REST API with a stable v1 and a v2 under development, Jira integration, webhooks, Slack, SSO via OIDC and SAML, and audit logs of every scan and remediation. It sits mid-scale because we found no public information on SCIM provisioning, granular roles per team, SARIF or CSV export of findings, or a self-hosted edition of the platform itself, though the CLI does run scans on customer-controlled infrastructure. 1 7 8 9 20 21 22

Report an error

The Engineering Manager

Code hosts cover GitHub, GitLab, Bitbucket and Azure pull checks plus Bitbucket Data Center and on-prem source code management, the REST API is documented in OpenAPI with stability levels stated, OIDC and SAML single sign-on come with the Teams plan, audit logging and webhooks exist, and the CLI runs scans on customer-controlled infrastructure. For administering six teams I found no public information on SCIM provisioning, SARIF export or granular roles per team, and cross-repository portfolio dashboards are only referenced generically as dashboard and reporting capabilities. 1 5 8 20 21 22

Report an error

The Platform Engineer

This fits my shop: Jenkins is in the supported CI list, merge-request comments work on GitLab Self-Managed and Bitbucket Data Center, single sign-on is SAML 2.0 and OpenID Connect with Azure AD named, and the captured pages explicitly document that CLI and customer-managed CI/CD scans run on customer-controlled infrastructure, so my 900 repositories can scan inside the Hanover network. The REST API is documented in OpenAPI for v1 and v2, with endpoints labelled stable, beta or experimental and a commitment not to deprecate v1 until v2 reaches parity — exactly the upgrade stability my pipelines need — and audit logging is available. We found no public information on SCIM provisioning, on SARIF export, or on a self-hosted edition of the platform itself, so findings management would still depend on the hosted service. 5 7 20 8 21 22 23 9

Report an error

The Product Security Officer

Self-managed code hosts including Bitbucket Data Center are documented with webhook-based triage workflows, the web API is published in OpenAPI format with stated stability levels, audit logging is available, and CLI scans run on customer-controlled infrastructure. We found no public information on SCIM provisioning, SARIF or CSV export of findings, or portfolio dashboards with trends across repositories. 5 9 20 21 22

Report an error

The Skeptic

On-prem and self-managed repositories are supported, scans can run on customer-controlled infrastructure with a Semgrep Network Broker mediating SCM access, Bitbucket Data Center pull request comments are documented down to the token permissions, and the REST API exists in two versions with per-endpoint maturity labels (Stable, Beta, Experimental) — the kind of candour I like in an API. We found no public information on SCIM provisioning, export of findings in SARIF or CSV, or granular roles per team, and portfolio reporting appears only as "Dashboard and reporting capabilities" with no trends shown. 5 8 9 18 20 21

Report an error

Source-code handling, security & compliance

Show reasoning
How this is scored

What the vendor does with the customer's source code and how it holds up in an audit: whether code leaves the customer's network or only results do, where and how long code and findings are stored, whether code or findings are used to train AI models and how to opt out, encryption, the vendor's own certifications (ISO 27001, SOC 2) with their scope, a stated availability or SLA and a public status page, audit logs, and reports mapped to OWASP, CWE, PCI DSS, ISO 27001 or the EU Cyber Resilience Act.

0 — Nothing on how source code is handled, secured or retained.

3 — A security page with general claims; code retention, AI training, certifications and the status page are basic or not described.

5 — How source code is accessed and whether it is stored is described, encryption in transit and at rest, SOC 2 or ISO 27001 stated, a public status page, a statement on whether code trains AI models, and compliance reports mapped to OWASP or CWE.

8 — Code retention periods stated, an option where code never leaves the customer's network, AI features documented with their model provider and an opt-out, certifications with scope and reports available on request, an SLA published, and reports mapped to several standards including PCI DSS or the Cyber Resilience Act.

10 — Source-code handling evidenced end to end: a data-flow diagram of what is sent where, zero retention options, independent audits published with scope, AI processing that can run on the customer's own models, and compliance evidence an auditor can export without asking the vendor.

Report an error

The Lead Developer

SOC 2 Type II with a stated audit period, a named pentest vendor, a public status page showing 99.98% uptime over 90 days, disclosed security incidents, and AI model providers named with a per-model opt-out and a custom provider option — and in CLI deployments the scans run on customer-controlled infrastructure. Two things hold it back for me: the terms grant a perpetual, sublicensable licence to use Customer Data to improve and develop Semgrep products and services, and we found no published SLA, no code retention period (the two-year cap covers personal information) and no PCI DSS or Cyber Resilience Act mapping. 9 10 12 13 18 22

Report an error

The Application Security Engineer

SOC 2 Type II is certified with the report period stated (December 1, 2023 to November 30, 2024) and downloadable from a trust-centre portal, dedicated pages map the product to ISO 27001:2022 Annex A controls and SOC 2 criteria with audit logs capturing every scan, finding and remediation with timestamps and user attribution, and the pages state plainly that Semgrep Inc. is not itself ISO 27001 certified. For a PCI DSS shop the gaps are concrete: we found no public information on PCI DSS or Cyber Resilience Act report mappings, no stated retention period for source code (two years is stated only for personal information), no published SLA, and the terms grant a perpetual licence to use customer data to improve and develop Semgrep products and services. 9 12 13 22 23

Report an error

The Engineering Manager

The trust centre shows a SOC 2 Type II report with its period stated, a full-scope whitebox penetration test by Include Security, a data processing agreement, a transfer impact assessment and GDPR documents, disclosed security incidents where no customer source code was accessed, audit logs, and CLI deployment that keeps code on my infrastructure. Semgrep Inc. is not itself ISO 27001 certified (its pages position the product as supporting a customer's certification), I found no published SLA, no encryption statement, no code retention period and no PCI DSS or Cyber Resilience Act mapping, and the captured terms grant the company a perpetual licence to use Customer Data to improve and develop its products — that last point goes to legal before anything is signed. 9 22 23 12 13 10

Report an error

The Platform Engineer

The vendor is SOC 2 Type II certified with a report covering December 2023 to November 2024, publishes audit logs recording every scan, finding and status change with timestamps and user attribution, runs an annual white-box penetration test by Include Security, and states plainly that Semgrep Inc. is not itself ISO 27001 certified while documenting how the product supports a customer's own ISO 27001 programme; there is a public status page showing 99.98% uptime over 90 days. The terms grant the company a perpetual licence to use Customer Data to improve and develop its products while the customer keeps all rights to Customer Data and Output, AI model preferences can be set per model with an opt-out, and secret validity checks run locally within our infrastructure. We found no public information on how long source code or findings themselves are retained, on a published SLA, or on PCI DSS or Cyber Resilience Act mappings. 22 23 9 13 12 10 3

Report an error

The Product Security Officer

A SOC 2 Type II report with its period stated, a trust centre offering the data processing agreement and a named penetration-test firm, an honest statement that the company itself is not ISO 27001 certified, compliance pages mapping capabilities to ISO 27001:2022 Annex A controls and SOC 2 criteria, audit logs with timestamps and user attribution, and a public status page at 99.98 percent uptime over 90 days. The gaps for an audit file: no stated retention period for source code (the two-year cap covers personal information), no published service level agreement, and no explicit statement on whether source code is used to train AI models — the terms grant a perpetual licence to use customer data to improve the products, and the documented opt-out covers which AI models are used rather than training. 9 10 12 13 22 23

Report an error

The Skeptic

Strong bones: SOC 2 Type II certified with the report period stated (December 1, 2023 to November 30, 2024), a named vendor (Include Security) behind a full-scope whitebox penetration test, two disclosed incidents with the statement that customer findings and source code were not accessed, CLI scans on customer-controlled infrastructure, and AI providers named (OpenAI as default for Multimodal, Amazon Bedrock for autofix) with a per-model opt-out. The caveats a buyer must weigh: we found no public information on code retention periods or a published SLA, Semgrep Inc. is stated as not itself ISO 27001 certified, and the terms grant the company a perpetual, sublicensable licence to create derivative works of Customer Data "to improve and develop Semgrep products and services" — read that clause against the opt-outs before assuming it covers nothing. 9 12 18 22 23

Report an error

European sovereignty panel opinion

Show reasoning
How this is scored

Who the contracting entity and the parent company are, where source code, findings, SBOMs, account and usage data are stored and processed, and who the subprocessors are (hosting, AI models used for autofix and triage, vulnerability data sources, email delivery, support and payment tools). Independently sourced by the sovereignty pipeline. Public vulnerability databases and package registries the scanner queries are part of what the product does and are not counted against it; where the customer's code and findings are stored and who processes them is. A self-hosted edition the customer runs on its own European servers counts as the customer's chain, whatever the vendor's seat.

0 — Non-EU vendor and contracting entity, hosting unstated, subprocessors unnamed, and no data processing agreement offered.

3 — A privacy policy and a data processing agreement exist, but hosting location is unstated or non-EU and the subprocessors are not listed.

5 — EU hosting available and an EU contracting entity or EU representative, but parts of the chain (AI models, support, payment) run through non-EU providers without a stated safeguard.

8 — EU hosting on named infrastructure, EU contracting entity, subprocessors listed with their locations, a data processing agreement published, and security certifications (ISO 27001, SOC 2) stated — or a self-hosted edition documented that keeps code and findings on the customer's own servers.

10 — Sovereign end to end and evidenced: European ownership, entity, hosting and every subprocessor including the AI models, or a self-hosted edition with no call home, data export and deletion documented, and certifications published with their scope rather than implied.

Report an error

The Lead Developer

A Delaware corporation governed by California law with San Francisco venue, servers stated to be located in the United States, and AI processing through OpenAI and AWS Bedrock — no EU hosting, EU contracting entity or EU representative appears in the captured pages. A data processing agreement, a data transfer impact assessment and a GDPR statement are published in the trust centre, and local CLI runs keep source code on our machines, but everything the platform hosts sits in the US. 9 12 13 23

Report an error

The Application Security Engineer

The contracting entity is Semgrep, Inc., a Delaware corporation governed by California law, with servers stated to be in the United States on AWS, WorkOS for authentication, OpenAI as the default Multimodal provider and AWS Bedrock for the Assistant — an entirely US chain for the hosted platform, softened by a published subprocessor list, a data processing agreement, a data transfer impact assessment and GDPR statements in the trust centre. The open-source CLI can scan locally without a login, which keeps code on my own servers, but the platform's findings, triage and AI features are US-hosted. 4 9 12 13 23

Report an error

The Engineering Manager

The chain is American end to end as documented: a Delaware corporation, California governing law and venue, servers located in the United States on Amazon Web Services, and the AI providers named are OpenAI and Amazon Bedrock. A data processing agreement and GDPR documents exist on the trust centre and the local CLI keeps source code on our own machines, but platform findings sit on US hosting; I found no public information on an EU hosting option, an EU contracting entity or an EU representative. 12 13 9 23 4

Report an error

The Platform Engineer

The contracting entity is Semgrep, Inc., a Delaware corporation at 799 Market Street, San Francisco, the servers sit in the United States on Amazon Web Services, California law governs with venue in San Francisco County, and named processors in the chain — WorkOS for authentication, OpenAI and AWS Bedrock for the AI features, Google Analytics on the website — are all US-based. A data processing agreement, GDPR materials and a subprocessor list exist in the trust centre, and the open-source Community Edition under LGPL 2.1 runs scans locally without a login, which keeps my code on my own servers in Hanover. We found no public information on an EU hosting option for the hosted findings platform, so everything beyond the local scanner runs through the US chain. 12 13 9 23 4

Report an error

The Product Security Officer

The contracting entity is a Delaware corporation under California law with servers in the United States on AWS, and the AI providers OpenAI and AWS Bedrock plus the authentication provider WorkOS are named, with a data processing agreement and a subprocessor list offered in the trust centre. We found no public information on EU hosting or an EU contracting entity or EU representative; the only chain we could keep in Europe is running CLI scans on our own customer-controlled infrastructure. 9 12 13 22

Report an error

The Skeptic

The documented chain is American throughout: Semgrep, Inc., a Delaware corporation under California law with exclusive venue in San Francisco County, servers stated to be located in the United States on AWS, and named processors WorkOS for authentication, OpenAI as the default Multimodal model provider, Amazon Bedrock, and Google Analytics. A data processing agreement, a GDPR listing and a subprocessor list do exist in the trust centre, and CLI scans on customer-controlled infrastructure can keep source code off Semgrep's servers. We found no public information on EU hosting, an EU contracting entity or EU residency for the platform where findings and triage are stored. 9 12 13 22 23

Report an error

Pricing transparency

Show reasoning
How this is scored

Whether a buyer can compute the real annual cost from public pages alone: price per developer, contributor, committer or line of code and per plan, how contributors are counted, what the free tier and open-source plan include, which scanners (SAST, SCA, secrets, containers) and features (SSO, self-hosting, AI autofix, compliance reports) cost extra, limits on repositories, scans or lines of code, minimum seats, monthly versus annual terms, and VAT.

0 — No public prices at all; every customer is a sales call.

3 — A starting price exists, but how developers or lines of code are counted and what the add-ons cost is unclear.

5 — Plans are public with their price per developer or per lines of code and main inclusions, but at least one commonly needed piece (SSO, self-hosting, an extra scanner, AI autofix or the enterprise tier) is unpriced or "contact sales".

8 — Every plan priced publicly with its counting rule (active contributors, lines of code) explained, scanners and add-ons priced, limits on repositories and scans stated, a free or open-source tier with its limits, and monthly and annual terms and VAT treatment stated.

10 — Complete price computability: annual cost derivable for a given number of developers, repositories and lines of code, with every scanner, add-on, self-hosted licence and support level priced and the cancellation and data-export terms published.

Report an error

The Lead Developer

The free tier is clear — 10 repositories, 10 contributors, 60 AI credits — and Teams prices every module publicly: Code at $30/month/contributor, Supply Chain at $30/month/contributor and Secrets at $15/month/contributor, with SSO included and repository limits per tier. Enterprise is volume pricing with a dedicated account manager and no figure, and we found no public information on how a contributor is counted, on annual versus monthly terms, or on VAT treatment — as an agency with rotating freelancers, that counting rule is the difference between a stable bill and a surprise. 8

Report an error

The Application Security Engineer

The pricing page publishes a free edition (up to 10 repositories, maximum 10 contributors) and Teams starting at $30/month per contributor with Code at $30/month/contributor, Supply Chain at $30/month/contributor and Secrets at $15/month/contributor, SSO included, repository limits per tier and AI credit allowances of 60 free and 20 per developer per month on Teams. I still cannot compute a real annual cost for 120 developers: the enterprise tier with volume pricing, on-prem source code management support and custom CI/CD carries no public price, and we found no public information on how a contributor is counted, annual versus monthly terms, VAT treatment, or the cost of AI credits beyond the allowance. 8

Report an error

The Engineering Manager

The free tier covers 10 repositories and 10 contributors, Teams is $30/month per contributor for Code and Supply Chain and $15/month per contributor for Secrets with up to 500 private repositories, AI credit allotments are stated per tier, and public repositories are unlimited on every plan. The Enterprise tier is volume pricing with no number, the counting rule for a contributor is not explained, and I found no public information on annual versus monthly terms, VAT treatment or what extra AI credits cost — so a full annual cost for my teams is not computable from the public pages. 8

Report an error

The Platform Engineer

Plans are public: the Free Edition scans up to 10 repositories with 10 contributors at $0, Teams starts at $30 / month per contributor with Code ($30/month/contributor), Supply Chain ($30/month/contributor) and Secrets ($15/month/contributor) as separately priced modules, SSO included, and 500 private repositories. The Enterprise tier is unpriced, offering unlimited repositories and volume pricing through a sales conversation. We found no public information on how a contributor is counted, on annual versus monthly terms, on VAT treatment, or on limits for scans or lines of code. 8 3

Report an error

The Product Security Officer

Free and Teams plans are priced publicly per contributor with per-module prices ($30/month/contributor for Code and Supply Chain, $15/month/contributor for Secrets), repository limits per tier, and included AI credits stated, plus a free tier limited to 10 repositories and 10 contributors. The enterprise tier carries no public price, how a contributor is counted is not explained, and we found no public information on monthly versus annual terms or VAT treatment. 8

Report an error

The Skeptic

The published parts are concrete: the free tier allows up to 10 repositories and a maximum of 10 contributors with Code and Supply Chain at $0/month/contributor and 60 AI credits, and Teams prices each scanner — Code ($30/month/contributor), Supply Chain ($30/month/contributor), Secrets ($15/month/contributor) — with private repository limits of 10, 500 and unlimited by tier and OIDC + SAML from Teams up. The gaps stop me computing a real annual cost: the Enterprise tier is "volume pricing" with no figure, "contributor" is never defined on the pricing page — an undefined contributor is exactly how every bot ends up on an invoice — and we found no public information on monthly versus annual terms or VAT treatment. 8

Report an error

European sovereignty — proven facts

5 of 6 dimensions proven

Built only from facts shown on the vendor's own pages. A dimension we could not prove is left open, not scored as zero.

Hosting location Outside the EU unverified
Why unverified?

The statement covers the website/Services servers; the CLI can also run scans locally so source code need not leave the customer's machine, but anything hosted by Semgrep is in the US.

0/4 pts 13 Report an error
Hosting options Self-hostable unverified
Why unverified?

Only the OSS Engine/CLI can be self-hosted; the Semgrep AppSec Platform itself is a SaaS-only offering, with Enterprise offering merely optional deployment in dedicated (Semgrep-run) infrastructure.

1/1 pts 12 Report an error
Open source Open source (copyleft) unverified
Why unverified?

The LGPL 2.1 applies to the Semgrep OSS Engine only; the same terms describe the Semgrep AppSec Platform Service and Pro rules as proprietary.

1/1 pts 12 Report an error
Data processing agreement On request unverified
Why unverified?

The DPA is only listed as a document inside the Trust Portal behind the 'Get access' / 'Ask for information' gate, so these excerpts show no self-serve way to view or conclude it online.

0/1 pts 9 Report an error
Transfer impact assessment Not determined unverified
Why unverified?

Not confirmed on the vendor’s own pages as captured.

— uncited Report an error

Questions about Semgrep

Is Semgrep European?

No. whats-best.ai classifies Semgrep as Rest of world. Its vendor, Semgrep, Inc., is incorporated in the United States, per the vendor’s page captured 6 Nov 2026.

Where does Semgrep store data?

Semgrep stores customer data outside the EU by default, per the vendor’s page captured 6 Nov 2026. The vendor’s pages mention the United States as hosting location, per the vendor’s page captured 6 Nov 2026.

Does Semgrep have a free plan?

Yes. Semgrep has a free plan, per the vendor’s page captured 6 Nov 2026.

What are the best alternatives to Semgrep?

As of 19 Nov 2026, the panel rates SonarQube (3.5/5), Aikido Security (3.5/5) and Endor Labs (3.5/5) highest among the other 18 Code Security & Quality (SAST/SCA) products evaluated. Ratings that look equal are rounded to half stars; the order uses the unrounded rating, then the name. The highest-rated European alternative is SonarQube (EU origin, foreign-owned, Switzerland). Ratings are the AI panel’s opinion, formed from public evidence: how scores are made. The full list, with European products first: Semgrep alternatives.

Where this could be wrong

What we left out

A claim that does not survive our checks costs us the claim, not the page. This is what was taken off this one.

Sources (23)

The pages every claim on this page was read from — each one checked, dated, and kept verifiable.

  1. 1 Vendor homepage semgrep.dev Checked 6 Nov 2026 Details →
  2. 2 Features page (SAST) semgrep.dev Checked 6 Nov 2026 Details →
  3. 3 Features page (secrets) semgrep.dev Checked 6 Nov 2026 Details →
  4. 4 Features page (open-source edition) semgrep.dev Checked 6 Nov 2026 Details →
  5. 5 Integrations page semgrep.dev Checked 6 Nov 2026 Details →
  6. 6 Documentation docs.semgrep.dev Checked 6 Nov 2026 Details →
  7. 7 API documentation docs.semgrep.dev Checked 6 Nov 2026 Details →
  8. 8 Pricing page semgrep.dev Checked 6 Nov 2026 Details →
  9. 9 Trust centre trust.semgrep.dev Checked 6 Nov 2026 Details →
  10. 10 Public status page status.semgrep.dev Checked 6 Nov 2026 Details →
  11. 11 About page semgrep.dev Checked 6 Nov 2026 Details →
  12. 12 Terms of service semgrep.dev Checked 6 Nov 2026 Details →
  13. 13 Privacy policy semgrep.dev Checked 6 Nov 2026 Details →
  14. 14 Code analysis (SAST & code quality) — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  15. 15 Code analysis (SAST & code quality) — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  16. 16 Dependencies (SCA), licences & secrets — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  17. 17 Dependencies (SCA), licences & secrets — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  18. 18 Developer workflow, quality gates & fixes — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  19. 19 Developer workflow, quality gates & fixes — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  20. 20 Platforms, deployment & administration — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  21. 21 Platforms, deployment & administration — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  22. 22 Source-code handling, security & compliance — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →
  23. 23 Source-code handling, security & compliance — found from sitemap docs.semgrep.dev Checked 6 Nov 2026 Details →