Semgrep

• GitHub, Snyk, or Checkmarx could bundle static analysis, dependency scanning, and AI triage into developer suites enterprises already buy, eroding the standalone budget line Semgrep sells against.
• Opengrep, which an outside reviewer records as having restored cross-function taint analysis, could close the remaining gap to the paid engine and recapture the open-source developers who feed Semgrep's paid funnel.
• Coding-agent vendors could ship their own security review, which would remove the distribution Guardian depends on inside Claude Code and Cursor.
• A public benchmark that measures Semgrep against rival scanners on the same code could undercut the accuracy Semgrep markets, since the 72 percent WebGoat detection rate in the reviewed record came from a benchmark Semgrep commissioned.
• Semgrep discloses no revenue, so a stall in commercial growth would leave the 204 million dollars raised without a public result to weigh it against.
• An intrusion that reached customer findings would strike directly at the trust a buyer extends to a vendor that reads source code.

Semgrep sells to the application security team and to the developers whose code it scans, and it frames the problem as scanner noise that buries the findings worth fixing. The platform unifies static analysis, dependency scanning, and secrets detection so the checks run where developers already work, in editors, pull requests, and build pipelines. The company traces that work back to 2017, when Drew Dennison, Isaac Evans, and Luke O'Malley founded it, and to 2020, when the team reignited the open source project that became the product.

The newer framing is code that AI agents write. Semgrep now sells Guardian, a plugin that installs into an AI coding agent and stops malicious packages, hardcoded credentials, and unsafe patterns at the moment the agent writes them.

Independent research quantifies the pain that vendor pages only assert. Lancaster University researchers testing four scanners on production code with known vulnerabilities found individual detection rates of 11.2 to 26.5 percent, and a separate academic evaluation measured a standalone Semgrep run on the OWASP Benchmark at an F1 score of 0.784. Gartner published a Magic Quadrant for Application Security Testing in October 2025, which is the category Semgrep sells into. [s5, s1, s8, s13, s14, s17]

Semgrep ships a free open-source engine and a paid platform on top of it. Semgrep Community Edition is the open-source program analysis engine, and its own repository warns that in security contexts the free edition misses many true positives because it analyzes within the boundaries of a single function or file. The paid platform adds cross-file analysis, Semgrep Supply Chain for reachability-aware dependency scanning and open-source malware blocking, and Semgrep Secrets, which uses semantic and entropy analysis and validates a credential by calling the service from inside the customer's own infrastructure.

An AI layer sits across the scanners. Semgrep Multimodal pairs deterministic static analysis with AI reasoning, and Semgrep reports that it finds up to 3.5 times more true positives at 19 percent lower cost per true positive than AI alone and cuts the findings a team must triage by 20 percent the day it is switched on. Triage decisions feed back, so the platform learns the organization-specific context that decides whether a finding is exploitable.

Semgrep Guardian carries the same scanning to AI coding agents. It bundles an MCP server, hooks, and skills into a plugin that installs into an agent, and its firewall blocks malicious packages at the network layer before they install.

The outside work in the reviewed record treats Semgrep as a configurable baseline rather than confirming its accuracy claims. Lancaster University researchers reconfigured it to reach 44.7 percent detection, more than combining four scanners. An independent preprint measured a standalone Semgrep run on the OWASP Benchmark at an F1 score of 0.784 and removed 496 of its false positives with an added filter. An outside reviewer records the 72 percent WebGoat result as vendor-funded research on a deliberately vulnerable training application rather than an independent cross-scanner test, and a public CVE for a Semgrep dependency is on record. [s6, s2, s3, s4, s8, s13, s14, s16, s15]

Semgrep competes against larger application security platforms and against a fork of its own engine. Semgrep's site offers a program for teams moving off Checkmarx or Snyk, so the company names the accounts it wants to take.

The structural pressure is on the open-source flank. Dark Reading reports that the triggering event for the split came on December 13, when Semgrep outlined changes to features of the free edition, and quotes an Opengrep backer saying Semgrep's biggest competitor had become its own open source engine. A consortium of application security companies then launched Opengrep, a fork of the Community Edition, keeping the same LGPL license and stating that it would restore the removed features and create an open source database of rules.

Parity between the fork and the paid platform is not established. An outside reviewer records that Opengrep restored cross-function taint analysis and other features Semgrep had moved behind its paid tier, while Opengrep's own page still lists cross-file analysis and extended language support among the capabilities it will unlock. The cited record shows no reproduction of the 20,000-plus proprietary rules the paid platform ships. [s1, s11, s12, s18, s16, s6]

Semgrep runs a product-led motion seeded by a free engine. The Free Edition costs nothing for up to 10 contributors and signs up directly, the Teams plan starts at 30 dollars per contributor per month, and Enterprise pricing is custom, so the published rate card covers the smaller team and the larger deal is negotiated.

The named-customer record is real and mostly vendor-sourced. Semgrep's repository names GitLab, Dropbox, Slack, Figma, Shopify, HashiCorp, Snowflake, and Trail of Bits as users, and an outside reviewer names Lyft, Dropbox, Snowflake, HashiCorp, Trail of Bits, GitLab, and Figma, six of them repeats of the repository's list. Scale is the company's own measurement, since SiliconANGLE reported that Semgrep disclosed adoption by hundreds of organizations since launch.

Analyst recognition arrived in 2025 and reached the record through Semgrep. The company announced its first placement in Gartner's Magic Quadrant for Application Security Testing, published in October 2025, in its own press release. Distribution now runs through AI coding tools as well, with Guardian downloadable from the Claude and Cursor marketplaces.

Semgrep discloses its funding and not its revenue. SiliconANGLE reported a 100 million dollar Series D led by Menlo Ventures that lifted outside funding to 204 million dollars, and no audited revenue figure appears in the reviewed sources. [s7, s6, s16, s10, s17, s8]

Semgrep's founders built the company from an open-source project into the platform it sells today. Drew Dennison, Isaac Evans, and Luke O'Malley founded Semgrep in 2017, and in 2020 the team reignited development of the open source project sgrep that became the product. Isaac Evans is chief executive.

The engine carries 16.4 thousand GitHub stars, the public registry holds more than 3,000 community rules by an outside reviewer's count, and an in-house security research team writes and maintains the 20,000-plus proprietary rules the paid platform ships.

Investors back the team. Menlo Ventures led the Series D, with Felicis Ventures, Harpoon Ventures, Lightspeed Venture Partners, Redpoint Ventures, and Sequoia Capital participating. [s5, s6, s16, s17, s10]

Semgrep publishes an inspectable trust portal, and an intrusion has now tested it. The portal announces a SOC 2 Type II report covering December 2023 through November 2024 and a full-scope whitebox penetration test performed by Include Security. The reviewed sources record no federal authorization and no product-specific mandate, so the assurance eases a procurement review without blocking a funded substitute.

The incident is the kind a code-scanning buyer fears. Semgrep disclosed that the Qilin group released Semgrep scan results of Semgrep's own source code, obtained through a Semgrep API token used to reach a sandbox environment, and says its completed post-incident investigations confirmed the threat actor did not access customer findings or source code and that the integrity of its source code and release processes was not compromised.

Semgrep's default keeps source code where the customer already holds it. Semgrep analyzes code locally or in the build environment and does not upload it by default, and its pricing page states that opting into AI-powered detection, triage, and remediation sends the part of the file containing a finding to a model for processing, with model vendors barred from training on the submitted code. [s9, s6, s7]

Add analyzed competitors to compare them side by side with Semgrep.

Semgrep sells to the application security team and to the developers whose code it scans, and its free tier extends that down to a single team. The Free Edition covers up to 10 contributors with authentication through GitHub or GitLab, so a team adopts without a sales conversation.

The free-to-paid funnel defines the segment more than any single named buyer. An outside reviewer describes the Community Edition as a free open-source engine limited to single-file analysis, and the AppSec Platform as the tier that adds cross-file analysis, dependency scanning, and secrets detection. The Free Edition sits between them, listing cross-file analysis with Pro rules while stopping at 10 contributors.

The newer segment is the team shipping AI-written code. Guardian installs into an AI coding agent and enforces the same rules there, which keeps the buyer inside the engineering organization rather than a separately cultivated market.

Semgrep's claimed advantage is pairing deterministic static analysis with AI reasoning rather than choosing one. Semgrep Multimodal uses rule-based analysis for classic flaws such as cross-site scripting and injection and AI models for business-logic vulnerabilities, and Semgrep reports 3.5 times more true positives at 19 percent lower cost per true positive than AI alone.

The scanning footprint spans the three established testing categories. Semgrep Code does cross-file analysis, Semgrep Supply Chain does reachability-aware dependency scanning and blocks open-source malware, and Semgrep Secrets uses semantic and entropy analysis and validates a credential inside the customer's own infrastructure, so one engine covers static analysis, composition analysis, and secrets.

The durable assets are the engine and the rule corpus, both Semgrep's own. The separate AI layer depends on outside model providers. The trust portal records that Semgrep Assistant added Amazon Web Services Bedrock as a model provider, and the pricing page states that opting into AI features sends the part of a file containing a finding to a model for processing.

Semgrep runs a product-led motion seeded by a free engine. The Community Edition and the Free Edition let teams adopt before they buy, and the engine carries 16.4 thousand GitHub stars.

The commercial proof is broad for this market and mostly vendor-sourced. Semgrep's repository names GitLab, Dropbox, Slack, Figma, Shopify, HashiCorp, Snowflake, and Trail of Bits as users, an outside reviewer repeats six of those names, and SiliconANGLE reported that Semgrep disclosed adoption by hundreds of organizations since launch. Semgrep announced its first placement in Gartner's Magic Quadrant for Application Security Testing itself, in a press release.

The company also competes for installed accounts directly. Semgrep's site offers a program for teams moving off Checkmarx or Snyk, and its repository lists Claude Code and Cursor plugins on those tools' official marketplaces, which puts the product where developers choose their agent tooling.

Semgrep publishes its rates, which is a positioning choice in this cluster. The Teams plan starts at 30 dollars per contributor per month, with Secrets at 15 dollars, while the Checkmarx page directs buyers to a custom quote and the Black Duck page to a no-obligation quote request.

The charged unit tells the buyer what Semgrep thinks it sells. A contributor is someone who made at least one commit to a scanned private repository in the past 90 days, so the meter tracks active developers rather than seats, assets, or scan volume.

Self-service stops at the free tier. The Free Edition costs nothing for up to 10 contributors and signs up directly, while Enterprise pricing is listed as custom, so the published rate covers the smaller team and the larger deal is negotiated.

Semgrep delivers software the customer runs, with a default that keeps source code in place. The engine analyzes code locally or in the build environment and does not upload it by default, so a team runs Semgrep in its own pipeline without sending source to a vendor cloud. The pricing page repeats that source code never leaves the machine or the CI environment when Semgrep runs locally or fully in CI.

The platform meets developers inside the tools they already use. Findings surface in editors and pull requests, and Guardian bundles an MCP server, hooks, and skills into an AI coding agent so a scan fires on every file write.

The AI features change what leaves the environment. Semgrep's pricing page states that opting into AI-powered detection, triage, and remediation sends the part of the file that has a finding in it to a model for processing, and that model vendors may not train on the submitted code.

Semgrep publishes an inspectable trust portal, and an intrusion has tested it. The portal announces a SOC 2 Type II report covering December 2023 through November 2024 and a full-scope whitebox penetration test performed by Include Security.

The incident is the kind a code-scanning buyer fears. Semgrep disclosed that the Qilin group released Semgrep scan results of Semgrep's own source code, obtained through a Semgrep API token used to reach a sandbox environment, and says its completed post-incident investigations confirmed the threat actor did not access customer findings or source code and that the integrity of its source code and release processes was not compromised.

A SOC 2 report and a third-party penetration test ease a procurement review without blocking a funded rival. The reviewed sources identify no federal authorization and no product-specific mandate, so a cautious buyer weighs the local-analysis default, the published reports, and the handling of the breach together.

Semgrep positions itself as the unified code-security platform built on an engine it gives away. The paid platform layers Code, Supply Chain, and Secrets over the open-source Community Edition, and the secrets page states that all products draw on the same underlying Pro and open-source engines, so a buyer consolidates three scanning categories instead of stitching point products together.

Outward, the platform reaches developers through the tools they already run. It integrates with editors, pipelines, and pull requests, it lets teams write rules in a pattern syntax that mirrors source code, and Guardian brings that scanning into AI coding agents through the Claude Code and Cursor marketplaces.

The structural tension is that the entry point is forkable. A consortium of application security companies launched Opengrep as a fork of the Community Edition after Semgrep changed the license on its open-source rules, and an outside reviewer records that the fork restored cross-function taint analysis and other features Semgrep had moved behind its paid tier. The cited record shows no reproduction of the 20,000-plus proprietary rules.

Semgrep's founders built the company from an open-source project into the platform it sells today. Drew Dennison, Isaac Evans, and Luke O'Malley founded Semgrep in 2017, and in 2020 the team reignited development of the open source project sgrep that became the product. Isaac Evans is chief executive.

The engine carries 16.4 thousand GitHub stars, the public registry holds more than 3,000 community rules by an outside reviewer's count, and an in-house security research team writes and maintains the 20,000-plus proprietary rules the paid platform ships.

Investors back the team. Menlo Ventures led the 100 million dollar Series D, with Felicis Ventures, Harpoon Ventures, Lightspeed Venture Partners, Redpoint Ventures, and Sequoia Capital participating.