Skip to main content
Glama

SafeWeave

Security scanner for AI-written code. SafeWeave runs inside your AI editor over MCP (Claude Code, Cursor, VS Code, Windsurf) and checks your code for vulnerabilities, leaked secrets, risky dependencies, IaC and container issues, so your assistant can fix findings in the same turn.

This repository is the open-core part of SafeWeave: the scanner engines (packages/scanners), the scan gateway (packages/gateway), shared types (packages/common), Docker files and the Claude Code skill. The MCP server itself ships on npm.

Quick Start

Run one command. It downloads the scanners and configures every editor it finds:

npx -y safeweave-mcp

Or add it to your editor's MCP config yourself:

{
  "mcpServers": {
    "safeweave": {
      "command": "npx",
      "args": ["-y", "safeweave-mcp"],
      "env": {
        "SAFEWEAVE_LICENSE_KEY": "your-license-key-here"
      }
    }
  }
}

The key is optional. Without one you get a 10-scan trial (SAST, secrets and dependencies). A free account at safeweave.dev/signup unlocks unlimited local scans with 7 scanners.

Related MCP server: Secure Code Review MCP Server

Plans

Plan

Price

Where scans run

Scanners

No key

Free

Local, 10-scan trial

3 (SAST, secrets, dependencies)

Free account

Free

Local

7 (everything except DAST)

Cloud

$29/mo

Local + hosted

All 8, plus PR checks, website scans and AI-builder rules

Cloud Plus

$59/mo

Local + hosted

All of Cloud, plus teammates

See safeweave.dev/pricing for details.

Alternative: Docker

If you prefer running via Docker:

git clone https://github.com/nickfluxk/safeweave.git
cd safeweave
cp .env.example .env   # Edit with your license key and SCAN_DIR
docker compose up

How Scanner Binaries Work

SafeWeave automatically downloads and manages the scanner binaries it needs on first use.

Tool

Version

Purpose

gitleaks

8.21.2

Secret and credential detection

trivy

0.58.0

Container vulnerability and IaC misconfiguration scanning

opengrep

1.16.3

SAST analysis (Semgrep-compatible, standalone binary)

  • Binaries are resolved in order: system PATH → ~/.safeweave/bin/ cache → auto-download

  • If you already have gitleaks or trivy installed via Homebrew/apt, those are used (never shadowed)

  • For SAST, system semgrep is preferred; if not found, opengrep (a standalone Semgrep-compatible fork) is downloaded — no Python required

  • Downloads happen in the background at startup — the server is ready immediately

  • Versions are pinned and updated with each SafeWeave release

  • If a download fails (no internet), that scanner is skipped gracefully

Supported platforms: macOS (arm64, x64), Linux (arm64, x64), Windows (x64)

Cache location:

~/.safeweave/
  bin/       # Downloaded binaries
  meta/      # Version metadata (JSON)

To force a re-download: rm -rf ~/.safeweave/bin ~/.safeweave/meta

What's Included

Scanner

What it detects

Plan

SAST

SQL injection, XSS, SSRF, path traversal, insecure crypto

Free

Secrets

API keys, tokens, passwords and private keys in source

Free

Dependencies

Known CVEs in npm, pip, Go and Cargo packages

Free

IaC

Terraform, Dockerfile and Kubernetes misconfigurations

Free account

Container

Vulnerabilities in container images

Free account

License

GPL, AGPL and other risky dependency licenses

Free account

Posture

Missing auth, rate limiting, CORS and security headers

Free account

DAST

Runtime API security issues in a running app

Cloud

MCP Integration

Works with any MCP-compatible editor — Cursor, Claude Code, VS Code, Windsurf, and more.

Once configured (see Quick Start above), ask your AI:

  • "Scan this project for security vulnerabilities"

  • "Check my dependencies for CVEs"

  • "Scan for hardcoded secrets"

  • "Run an IaC misconfiguration scan"

If running via Docker, use the SSE transport instead:

{
  "mcpServers": {
    "safeweave": {
      "url": "http://localhost:9000/sse"
    }
  }
}

HTTP API

Health Check

curl http://localhost:9000/api/health

Run a Scan

curl -X POST http://localhost:9000/api/scan \
  -H 'Content-Type: application/json' \
  -d '{"directory": "/path/to/project"}'

Scan Specific Scanners

curl -X POST http://localhost:9000/api/scan \
  -H 'Content-Type: application/json' \
  -d '{"directory": "/path/to/project", "scanners": ["sast", "secrets"]}'

Compliance Profiles

Built-in profiles: standard, hardened, owasp, soc2, pci-dss, hipaa

Set via MCP set_profile tool or create a custom .safeweave/profile.yaml in your project.

Configuration

Environment Variables

Variable

Required

Description

SAFEWEAVE_LICENSE_KEY

No

Your license key from safeweave.dev. Without it: a 10-scan trial

SCAN_DIR

Docker only

Host directory to mount for scanning

SAFEWEAVE_LICENSE_URL

No

License server URL (defaults to https://license.safeweave.dev)

Claude Code Skill: Auto-Scan Before Push

SafeWeave includes a Claude Code skill that automatically scans your code before every git commit, git push, or PR creation.

Install the skill

Copy the skill to your Claude Code skills directory:

mkdir -p ~/.claude/skills/secure-before-push
curl -sL https://raw.githubusercontent.com/nickfluxk/safeweave/main/skills/secure-before-push/SKILL.md \
  -o ~/.claude/skills/secure-before-push/SKILL.md

Or clone and copy:

git clone https://github.com/nickfluxk/safeweave.git /tmp/safeweave
cp -r /tmp/safeweave/skills/secure-before-push ~/.claude/skills/

What happens after install

Once installed, Claude Code automatically runs a SafeWeave scan whenever you say "commit", "push", "ship it", or "create a PR":

Severity

Action

Blocks push?

Critical / High

Fix immediately, re-scan

Yes

Medium

Warns, asks confirmation

User decides

Low / Info

No action

No

Prerequisites

SafeWeave MCP server must be configured. Run npx safeweave-mcp to set it up automatically.

Verify

Start a new Claude Code session and ask it to commit code. You should see a SafeWeave scan run before the git operation.

License

MIT. See LICENSE.

Related MCP Connectors

Related MCP Servers

  • F
    license
    Not graded
    quality
    D
    maintenance
    Local MCP server that scans code for security issues (secrets, dependencies, configurations, risky patterns) and integrates with GitHub Copilot in VS Code for automated pre-commit reviews.
    -
  • A
    license
    Not graded
    quality
    A
    maintenance
    Enables AI coding tools to scan projects for security vulnerabilities, hardcoded secrets, injection flaws, and privacy violations with 699 rules and 76 MCP tools, all running locally with zero telemetry.
    6 npm
    7
    MIT
  • A
    license
    Not graded
    quality
    B
    maintenance
    Enables MCP-compatible AI agents to scan code for leaked secrets, copyleft licenses, unprotected routes, missing privacy policies, and risky card handling before committing or shipping.
    81 npm
    MIT