Skip to main content

Security Review MCP: a finding reported in the code

Security Review MCP

licence Python Angular ci::status pypi::version docker::pulls documentation::badge

MCP server for security-oriented code review (vulnerabilities and misconfigurations), usable from VS Code / Copilot, Claude or a CI/CD agent. It provides review prompts, checklists, a pattern scan with focused Tree-sitter checks for request-to-SQL flows (Python, Java, Go, C#, Rust, Node.js), Angular DOM flows and unchecked message origins, an offline dependency analysis and a report database (SQLite or PostgreSQL) usable in pipelines.

📖 Full documentation: https://cyr-ius.github.io/security-review/

🇫🇷 Version française

Capabilities

Prompts security-review, api-security-review, configuration-review, codebase-review, dependency-review, triage-report, development-standards
Tools checklists, scan_code, scan_dependencies, saving and reading of reports with their checklist coverage (grouped by a normalized project name), status proposals (propose_finding_status: an agent never closes a finding), project posture and gate, SARIF export, trends, rule management
CLI security-review-mcp (MCP server: serve, local for the local mode) and security-review-controller (serve, db, advisories, rules, token, user, audit), exit codes 0 / 1 / 2
Pipelines security-review-ci (gate, sarif, trend, scan-code, scan-deps, which store their reports through the API): lightweight client of the REST API for Jenkins, GitLab or GitHub, with a ci token restricted to the selected projects, no database driver, installed alone without Python (uv tool install security-review-ci, or from the Controller server without PyPI); on a workstation (commit hook), with a personal ci token; agent-review: review of each commit by the coding agent of the developer (Claude Code, Copilot or Codex)
Rules More than 200 rules and more than 120 checklist items: Python, Java, Go, C#, Rust, TypeScript / Node.js backend, Angular, Docker, Kubernetes, configuration, GitLab CI, Jenkins, Spring, FastAPI, prompt injection aimed at AI assistants (hidden characters, injected instructions; read guard security-review-ci guard, a hook for Claude Code, Copilot or Codex), copied into the database at first start, plus your own rules and checklists (TOML import/export from the CLI or the web interface)
Development standards conventions by technology managed in the web interface (TOML import/export) and handed to the AI assistant before it codes (prompt, get_development_standards / list_development_standards tools, resources)
Vulnerabilities manifests, lockfiles and CycloneDX SBOMs checked against an OSV database (PyPI, npm, Maven, Go, crates.io, NuGet) imported offline or updated every day by the Controller server, optional proxy
Transport HTTP (shared mode) or stdio (local mode), with Bearer tokens stored in the database (token CLI) with expiry and rotation, and TLS (optionally mutual)
Audit and metrics audit log of triage decisions, tokens and project requests, and Prometheus metrics on /metrics (metrics token)
Connectors findings sent to other tools: Jira and ServiceNow built in, third-party connectors for any other system (vulnerability register, GRC...); each connector switched on in the web interface, by hand or automatically above a severity threshold, one record per issue
Deployment MCP server and Controller server (REST API, web interface) separated on a shared database, Alembic migrations, Docker Compose and Kubernetes examples
Enrollment an MCP client without a token is told to create a personal MCP token in the web interface; then the request_project MCP tool creates the project of the review, or requests it (or the access to an existing project) for an administrator to approve, which gives the user access to it; a refused token gets a 401 that says how to get a new one
OAuth login optional, turned on from the web interface: the MCP host needs no token, it signs in through the browser (OAuth 2.1 with PKCE, the Controller server as authorization server) and acts as a personal MCP token of the user
Web interface in English, French or Spanish: dashboard, triage by a person (approver of the accepted risks and false positives, reason and expiry of the accepted risks), risk acceptances to renew or remove, with append-only comments, per-project list of open findings for an expert review, management of rules, checklists, tokens (per project or for all projects), general settings (report language, Controller server address), outgoing proxy (OSV and OIDC) and OSV updates, local or OIDC user accounts, the only ones to open the interface (tokens are for machines: MCP agents, pipelines, scripts; personal or role API tokens for the REST API and Swagger, personal MCP tokens on the projects of the user, unless SECREVIEW_PERSONAL_TOKENS_ENABLED=false; initial admin account in the logs), optional OpenID Connect OIDC login, configurable in the interface, with accounts created on the fly, and roles and permissions (roles on the whole server and projects of each account with their own per-project permissions, project search and multiple selection, and OIDC mappings of the identity provider groups)

Details: how a review works (rules, checklists, standards, AGENTS.md), prompts and tools, whole codebase analysis, dependencies, gate, SARIF, trends.

Quick start

The product runs as two services sharing one database: the MCP server (security-review-mcp), used by the agents and the CI, and the Controller server (security-review-controller), used by people: web interface, REST API, tokens and accounts. See two services, two roles.

uv venv && source .venv/bin/activate
uv pip install security-review         # Python ≥ 3.14; [postgres] extra for PostgreSQL

# from the same directory (shared SQLite file)
security-review-controller db upgrade                              # schema (Alembic)
security-review-controller user create admin --role admin --all-projects   # web interface account
security-review-controller token create --label alice              # token of the agent
security-review-controller serve --port 8080                       # /ui interface and REST API
security-review-mcp serve --port 8000                              # /mcp endpoint (streamable HTTP)

security-review-ci gate --url http://localhost:8080 --project portal  # block a pipeline (ci token: SECREVIEW_CI_TOKEN)

Host configuration (VS Code with a token, .vscode/mcp.json; Claude and .mcp.json: see connecting the host):

{
  "servers": {
    "security-review": {
      "type": "http",
      "url": "http://127.0.0.1:8000/mcp",
      "headers": { "Authorization": "Bearer ${input:secreview-token}" }
    }
  },
  "inputs": [
    { "id": "secreview-token", "type": "promptString", "description": "Token", "password": true }
  ]
}

Step by step: deploying the server, connecting the host (VS Code, Claude, enrollment). Other cases: offline installation, local mode (stdio), Bearer tokens, TLS, Docker images (ghcr.io/cyr-ius/security-review, docker.io/cyrius44/security-review) and Docker Compose with PostgreSQL, architecture and Kubernetes, SECREVIEW_* variables.

Limits

The scan produces leads, with false positives. Its Tree-sitter checks cover only simple local flows inside one file. It replaces neither a full SAST or SCA analysis, nor a human review. See security and limits.

Contributing

uv sync --extra dev && uv run pytest && uv run ruff check .

Development happens on dev, and a version is released by pushing a vX.Y.Z tag on master. See the contribution guide and AGENTS.md (instructions for AI agents).

Security and license

Report a vulnerability privately, as described in SECURITY.md. The project is distributed under the Apache 2.0 license.

Metadata

Release files for security-review 10.0.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for security-review 10.0.0
File Size Uploaded Provenance Actions
security_review-10.0.0.tar.gz 1.7 MB With provenance Details for security_review-10.0.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for security-review 10.0.0
File Interpreter ABI Platform Actions
security_review-10.0.0-py3-none-any.whl Python 3 none any Details for security_review-10.0.0-py3-none-any.whl

Total release size: 3.4 MB

Release files / security_review-10.0.0.tar.gz

Download URL security_review-10.0.0.tar.gz
Size 1.7 MB
Tags Source
SHA-256 checksum
How to use checksums
c36ba396bc7bf135ee1aa3b45823823f9e888f3fea8fc77ac02704cda03d63ca
BLAKE2b-256 checksum
How to use checksums
190d86cb0d26bd658edc4ba95931b48746a0f5d576314b8ac5010f92e271b3f4
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 30, 2026.

Transparency log

Release files / security_review-10.0.0-py3-none-any.whl

Download URL security_review-10.0.0-py3-none-any.whl
Size 1.6 MB
Tags Python 3
SHA-256 checksum
How to use checksums
fcb3f867b7ebec5e32018e3d2c193b2bcec466b888d66be423f22713cccdb666
BLAKE2b-256 checksum
How to use checksums
bb4d674e1b9d1c89086b312693eb037b077eeb2a7b0c8ab344756846cf13c68a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 30, 2026.

Transparency log

Security for security-review 10.0.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 2 files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for security_review-10.0.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 30, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page