Semgrep
Also known as: semgrep.dev, Semgrep Inc., r2c, Semgrep Multimodal, Semgrep Assistant, Semgrep Guardian, Semgrep MCP server, Semgrep AppSec Platform
Semgrep is an application security platform for code, dependencies and secrets whose AI layer, Multimodal, finds logic flaws, triages findings with memory of past decisions and opens draft fix pull requests through Autofix. Guardian scans code inside Cursor, Claude Code and other coding agents as it is written, and the Semgrep MCP server lets any agent triage, fix and file tickets. Teams starts at $30 per contributor a month per product, with a free plan for 10 contributors.
Semgrep, the San Francisco application security company formerly known as r2c, scans code for vulnerabilities, risky dependencies and leaked secrets, for developers and for the coding agents they run. Lyft, Slack, Dropbox, GitLab, HashiCorp, Vanta, Squarespace, Thinkific, Acrisure and Trail of Bits are among its users. Teams see a 60% backlog cut on day one, 80% fewer false positives to triage across SAST and SCA and 3.5 times more true positives than AI alone, with 96% agreement on its AI triage, and 95% of security reviewers validated its AI findings across more than 6 million of them.
Semgrep Multimodal, the AI layer renamed from Assistant in March 2026, pairs static analysis with model reasoning to find business logic flaws such as broken authorization, triages each finding with an explanation, holds back pull request comments it judges false positives for security staff to review, tags findings by function such as payments or login, and learns from memories, the team's past triage decisions and custom instructions. Autofix writes the code change for Code and Supply Chain findings, opens a draft pull or merge request with an AI written description and can group several findings from one rule into one change.
Agentic Workflows investigate and remediate across a codebase on a schedule or on demand. Guardian plugs into Cursor, Claude Code, GitHub Copilot, Windsurf and Kiro through an MCP server, hooks and skills to scan every file an agent writes against company policy before it reaches a commit, and the Semgrep MCP server, in beta since November 2026, lets any agent triage findings, start fixes, retry failed scans and file Jira or Linear tickets within the user's own permissions.
Scans run in the command line, in a customer's own pipelines, or as managed scans on Semgrep's SOC 2 Type II certified AWS infrastructure. Enterprise adds on premises source control, optional dedicated infrastructure and a custom AI model provider, and by default Multimodal runs on OpenAI with automatic fallback to Amazon Bedrock. Teams get OIDC and SAML single sign on and role based access. The Free plan covers up to 10 contributors with Code and Supply Chain and 60 AI credits; Teams starts at $30 per contributor a month for each of Code and Supply Chain and $15 for Secrets, with 20 AI credits per developer, and Enterprise is quoted with 50.
Semgrep serves application security teams that want triage and fixes delivered as reviewable pull requests, and engineering teams whose coding agents need security checks while they write. It does not drive a browser, and its AI models are chosen by Semgrep unless an Enterprise customer brings its own provider.
Vendor details
Canonical URL
https://semgrep.dev
Category
Security / SOC agent
Subcategory
Application security platform whose AI triages findings with memory, opens draft fix pull requests, runs agentic remediation workflows and secures code inside coding agents through Guardian and an MCP server
Funding status
Venture backed private company in San Francisco (Semgrep Inc., formerly r2c).
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
GitHub, GitLab, Bitbucket and Azure DevOps pull request checks and Autofix requests; Jira and Linear tickets; Slack and Jira for malware incidents; VS Code and JetBrains; Guardian for Cursor, Claude Code, GitHub Copilot, Windsurf and Kiro; Semgrep MCP server for Claude, ChatGPT, Codex, Cursor and VS Code; REST API and webhooks; Wiz, Cortex, Palo Alto Networks, Sysdig and StackHawk.
In practice
An application security team turns on Multimodal across 400 repositories. Day one triage marks a large share of old findings as likely false positives with explanations, the team agrees or overrides each on the pre production page, and those decisions become memories for later scans.
A developer's pull request adds an endpoint that skips an ownership check. Semgrep flags the broken authorization in a comment, and Autofix opens a draft fix request that adds the check, which the developer reviews and merges.
An engineering team installs Guardian in Claude Code and Cursor, so each file an agent writes is scanned against the company's policies and fixed before commit, while the security lead uses the MCP server to file Linear tickets for findings that need a person.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
12.0 / 14 capabilities · 86%
| Integrations & Tool Calling | Full |
|---|---|
|
Findings flow into GitHub, GitLab, Bitbucket and Azure DevOps as pull request comments and checks, and Autofix opens draft pull or merge requests through the Semgrep GitHub App. Tickets go to Jira and Linear, malware incidents route through Slack and Jira, IDE plugins cover VS Code and JetBrains, and Wiz, Cortex, Palo Alto Networks, Sysdig and StackHawk add cloud context. Sourcesemgrep.devread 2026-11-19 |
|
| Workflow Orchestration | Full |
|
Autofix runs a full fix cycle on its own: it generates the change, creates a branch, applies it, opens a draft request with a written description and can group several findings from one rule into a single change. Agentic Workflows investigate and remediate vulnerabilities across a codebase as scheduled or admin triggered jobs, with status and task logs for each run. Sourcedocs.semgrep.dev/semgrep-code/triage-remediation/autofix.mdread 2026-11-19 |
|
| Knowledge Grounding & RAG | Full |
|
Triage and fixes are grounded in the code itself: cross file dataflow from the Pro engine, Mandoline's reachability tracing of control and data flow, component tags that mark findings touching payments or authentication, and the organization's own rules and past triage decisions. The AI works on the repository's real context rather than a generic description of the vulnerability class. Sourcedocs.semgrep.dev/semgrep-multimodal/overview.mdread 2026-11-19 |
|
| Human Oversight & Guardrails | Full |
|
Autofix opens every change as a draft that a developer reviews and merges, and noise filtering holds back comments rather than deleting them, listing each decision on a pre production page where security staff agree or disagree. Through the MCP server an agent can do only what its user's role and token allow, read only tokens cannot open fix requests, and only admins trigger workflows. Sourcedocs.semgrep.dev/semgrep-code/triage-remediation/autofix.mdread 2026-11-19 |
|
| Security, Identity & Governance | Full |
|
Semgrep Inc. holds a SOC 2 Type II report and backs it with outside penetration testing, and managed scans run on SOC 2 Type II certified AWS infrastructure. Teams and Enterprise plans add OIDC and SAML single sign on and role based access control, and MCP access signs in through OAuth or scoped tokens, including a read only token and an agent token limited to scan operations. Sourcedocs.semgrep.dev/compliance/soc2read 2026-11-19 |
|
| Observability & Auditability | Full |
|
Each finding keeps an activity record showing triage, fix requests and who acted, and the pre production page lists every comment noise filtering withheld and why. Workflow runs expose job status, task logs and credit usage, reports and weekly priority emails summarize risk, and the platform tracks how often reviewers agree with the AI's triage, 96% across customers. Sourcedocs.semgrep.dev/semgrep-multimodal/overview.mdread 2026-11-19 |
|
| Memory & State Persistence | Full |
|
Memories turn a team's triage decisions into reusable rules, so a false positive dismissed once stays suppressed across later scans, and custom instructions per project and per rule shape remediation guidance to house standards. Agents working through the MCP server reuse the organization's triage memories, so a coding assistant inherits the security team's past judgment. Sourcesemgrep.devread 2026-11-19 |
|
| Deployment & Data Residency | Full |
|
Scans can run entirely on a customer's own machines through the open source command line tool or self managed pipelines, or as managed scans on Semgrep's AWS infrastructure. Enterprise supports on premises source code management, optional dedicated infrastructure and a custom AI model provider, so code sent for AI analysis can stay with a provider the customer chooses. Sourcesemgrep.dev/pricingread 2026-11-19 |
|
| Prebuilt Agents / Templates / Packs | Partial |
|
The rule registry and Pro rules give teams ready detection for many languages and frameworks, and Guardian ships as a packaged plugin with an MCP server, hooks and skills for popular coding agents. Agentic Workflows, Autofix and triage are features configured per organization rather than a catalog of finished security agents a team picks and deploys. Sourcesemgrep.dev/blog/2026/introducing-semgrep-guardian-real-time-security-for-ai-written-coderead 2026-11-19 |
|
| Triggers & Channel Coverage | Full |
|
Scans run on every pull or merge request, in CI and on a schedule, and Guardian's hooks fire each time a coding agent writes a file. A malware scan starts within 30 minutes of a malicious package being discovered, admins schedule Agentic Workflows, and results arrive as pull request comments, platform findings and weekly priority emails to org admins. Sourcesemgrep.devread 2026-11-19 |
|
| Model Flexibility & Routing | Full |
|
Multimodal ranks models and selects among the providers a customer enables, using OpenAI first in most cases with automatic fallback to Amazon Bedrock, and Autofix needs Bedrock or Anthropic Claude models accepted. Enterprise customers can plug in a custom AI model provider, while the MCP server works with whichever agent and model a developer already runs. Sourcedocs.semgrep.dev/semgrep-multimodal/overview.mdread 2026-11-19 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
The Semgrep MCP server, a remote server in beta since November 2026, exposes ten areas of the platform, from triage and fixes to scans, tickets, rules, reports and workflows, to Claude, ChatGPT, Codex, Cursor and VS Code. A REST API and webhooks cover the platform, the command line tool is open source, and Guardian bundles its own MCP server, hooks and skills. Sourcesemgrep.dev/blog/2026/semgrep-mcp-serverread 2026-11-19 |
|
| Testing, Debugging & Optimization | Partial |
|
Security teams can review each noise filtering decision and record agreement or disagreement, and the platform reports how often reviewers agree with AI triage, which tells them how far to trust it. There is no harness for testing a change to memories, instructions or model settings against a fixed set of known findings before it applies across every scan. Sourcedocs.semgrep.dev/semgrep-multimodal/overview.mdread 2026-11-19 |
|
| Browser / Computer-use | Not documented |
|
Semgrep works by parsing code, calling repository and ticketing APIs and plugging into coding agents through MCP and hooks. None of its agents operates a browser or a desktop, so checks that need a running application, a signed in session or a click through a user interface fall to dynamic application testing tools such as those from its partner StackHawk. Sourcesemgrep.devread 2026-11-19 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Free up to 10 contributors; Teams from $30 per contributor/mo per product (Secrets $15); Enterprise custom
Per contributor per month for each product (Code, Supply Chain, Secrets), with monthly AI credits per developer
Included quota
Free: 60 AI credits, 10 private repos. Teams: 20 AI credits per developer a month, up to 500 private repos. Enterprise: 50 AI credits per developer a month, unlimited repos
What is public
Semgrep is free for up to 10 contributors with Code and Supply Chain. Teams costs $30 per contributor a month for Code, $30 for Supply Chain and $15 for Secrets, and Enterprise is quoted.
Billing mechanics
Each product bills per contributor per month, and AI features draw on monthly credits: 60 on Free, 20 per developer on Teams and 50 per developer on Enterprise.
Cost watchouts
Each product is priced separately per contributor, so Code, Supply Chain and Secrets together come to $75 per contributor a month on Teams.
Variable cost rationale
Seats are per contributor per product, and AI features draw on a monthly credit allowance per developer, with scans and workflows run by Semgrep using credits.
Additional watchouts
A custom AI model provider and self managed repositories come only with Enterprise.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free plan for up to 10 contributors and 10 private repositories with Code and Supply Chain and 60 AI credits; free trial of paid plans
Lowest paid plan
Teams, from $30 per contributor a month (Code or Supply Chain)
Commercial notes
Enterprise adds on premises source control, dedicated infrastructure, a custom AI model provider and volume pricing, and Guardian is free for 30 days.
Key ambiguities
What AI credits cost beyond the monthly allowance is not public.
Support SLA / resale
Enterprise adds a dedicated account manager and tailored onboarding.
Missing data
The Enterprise price and any rate for AI credits past the monthly allowance are not public.
Related vendors
- 7AI — Agentic SOC from the Cybereason founders: domain specialized agents…
- Abnormal AI — Behavioral AI email security with three named autonomous agents: a…
- AirMDR — AI native MDR whose AI analyst investigates every alert and writes…
- Airrived — Agentic OS that unifies SOC, GRC, IAM, vulnerability management, IT,…
- Andesite — Human-AI SOC platform where customers build their own agents and…
- Anomali — Anomali combines a security data lake, ThreatStream intelligence and…
Alternatives to Semgrep
Other products that do the work Semgrep does, and what each one is.
- ContraForce
- CrowdStrikeCybersecurity platform
- D3 Security (Morpheus)
- TinesWorkflow and AI orchestration platform
- Torq
- Dropzone AI