ARGUS · SAST

SAST that finds logic flaws, not just pattern matches.

ARGUS runs static application security testing on every scan: an open rule engine for known patterns, then AI passes that follow data across files and look for the flaws rules cannot express, like a missing ownership check. Every AI finding is challenged before it reaches you.

Coverage

What it looks for

Rule-based SAST runs on every scan, with coverage that follows the open Semgrep registry. With AI scanning on, four extra passes run on the languages below.

  1. 01

    Taint analysis

    Untrusted input followed from source to sink across files. Python, JavaScript, TypeScript, Ruby, Go, Java and PHP.

  2. 02

    Business logic

    Insecure direct object references, privilege escalation, auth bypass and broken sequences. Python, JavaScript, TypeScript and Ruby.

  3. 03

    LLM application risks

    Prompt injection, insecure output handling and excessive agency in code that calls an LLM SDK, mapped to the OWASP Top 10 for LLM applications.

  4. 04

    Race conditions

    Time-of-check to time-of-use bugs, double spends and transaction-scope errors.

How it works

From scan to fix

  1. 01

    Scan a repository

    Connect GitHub or a public Git URL. The code is cloned to a temporary folder and deleted when the scan ends.

  2. 02

    Rules, then AI passes

    Rule findings are scored and mapped to MITRE ATT&CK where a technique fits. AI passes add what rules miss.

  3. 03

    Every AI finding is challenged

    Independent reviewers try to refute each AI finding. If the evidence does not hold up, it is dropped.

  4. 04

    Read the evidence, then fix

    Each finding shows the reasoning, an attack scenario and the taint chain file by file. Assign it to the fix agent for a pull request.

Why teams use it

Fewer, better findings

Verified, not guessed

Adversarial verification removes AI findings that cannot be substantiated, so you are not triaging hallucinations.

Ranked with everything else

SAST findings share one 0 to 100 priority with dependencies, secrets and infrastructure, so the list is one list.

Deep scan when it matters

A Deep Agentic scan lets an AI agent follow its own hypotheses across the repository, with the same verification.

Limits, stated plainly

What it does not do yet

We would rather you hear it here than find it in week two.

  • AI passes run only when AI scanning is turned on, and only for the languages listed above.
  • Rule coverage follows the open rule registry rather than a fixed language list.
  • Turning a rule off stops new findings from the next scan; existing findings stay open until you close them.
  • Test files are excluded from SAST by default.
FAQ

SAST questions, answered

Straight answers about SAST and how ARGUS does it. Updated 17 November 2026.

Still have questions? Contact us

What is SAST?

SAST (static application security testing) analyses source code without running it, to find vulnerabilities such as injection, unsafe deserialization and broken access control early in development. ARGUS runs rule-based SAST on every scan, then adds AI passes for flaws rules cannot express, such as a missing ownership check.

What is a SAST scan?

A SAST scan is one run of a static analyser over your codebase. It produces findings with a file, line, severity and suggested fix. In ARGUS, SAST runs as part of every repository scan, and its findings share one ranked list with dependency, secret and infrastructure findings.

What is the difference between SAST and DAST?

SAST reads your source code from the inside before it runs. DAST (dynamic application security testing) attacks the running application from the outside by sending requests. SAST pinpoints the line to fix early; DAST finds issues that only appear at runtime. ARGUS focuses on code-level analysis: SAST, SCA, secrets and infrastructure config.

What is the difference between SAST, SCA and DAST?

SAST checks the code you write. SCA (software composition analysis) checks the open-source packages you import. DAST tests the running application from outside. ARGUS covers SAST and SCA in the same scan, together with secrets and infrastructure-as-code checks.

Can SAST find business logic vulnerabilities?

Rule-based SAST rarely can, because logic flaws depend on what your code is meant to do. ARGUS adds an AI business-logic pass that looks for insecure direct object references, privilege escalation, auth bypass and broken sequences in Python, JavaScript, TypeScript and Ruby.

Why do SAST tools produce so many false positives?

Pattern rules match code shapes without knowing whether input is trusted or a path is reachable. ARGUS reduces the noise in three ways: every AI finding is challenged by independent reviewers and dropped if unsupported, findings are ranked by real risk, and admins can turn off rules that do not fit.

Which languages does ARGUS SAST support?

Rule-based coverage follows the open Semgrep registry rather than a fixed list. The AI passes cover Python, JavaScript and TypeScript for all four checks, with Ruby, Go, Java and PHP covered by some of them, as listed in the coverage section above.

What is a Deep Agentic scan?

A Deep Agentic scan lets an AI agent explore your repository, follow its own hypotheses across files and report only what it can substantiate. Its findings go through the same adversarial verification as other AI findings. It uses more credits than a standard scan, so teams run it when they want depth.

Does ARGUS store my source code?

No. ARGUS clones the repository into a temporary folder for the scan and deletes it when the scan finishes, including when a scan fails or times out. Findings are kept; the code is not.

Can SAST run on pull requests?

Yes. The ARGUS pull request review runs rule-based checks scoped to the changed lines plus secret detection, with an optional AI security pass. Findings appear as inline comments, and a commit status can block the merge.

Try it on one repository.

Connect a GitHub repo and see the findings, ranked, in your first scan.