Powered by GLM-5.2

AgentGuard

Rules find suspects. GLM-5.2 confirms, fixes, and verifies.

Scan your Python and JavaScript code for security vulnerabilities. GLM-5.2 reviews every finding, filters false positives, and auto-fixes the real bugs — all in one command.

Try Pro Free 14 Days GitHub
pip install agentguardp

GLM-5.2 vs Claude Code

Independent security benchmark — lower cost, higher accuracy

ModelF1 Score (Vuln Detection)Cost per Finding
Claude Code32-37%~$1.02
GLM-5.239%$0.17

Source: Semgrep security evaluation, via cybernexora.com and paddo.dev

See it in action

scan to GLM-5.2 verify to auto-fix to self-validate

agentguard pipeline ./src --glm-fix
$ agentguard pipeline ./src --glm-fix Scanning 47 files... [CRITICAL] SQL Injection — db.py:23 [HIGH] Hardcoded Password — config.py:8 [MEDIUM] eval() usage — utils.py:45 Found 3 issues. Sending to GLM-5.2 for review... GLM-5.2 confirmed: 3/3 are real vulnerabilities. Auto-fixing with GLM-5.2... Fix 1/3: SQL Injection to Parameterized query - cursor.execute("SELECT * FROM users WHERE id=" + user_id) + cursor.execute("SELECT * FROM users WHERE id=%s", (user_id,)) Fix 2/3: Hardcoded Password to Environment variable - PASSWORD = "admin123" + PASSWORD = os.environ.get("DB_PASSWORD") Fix 3/3: eval() to ast.literal_eval - data = eval(user_input) + import ast; data = ast.literal_eval(user_input) Re-scanning fixed code... All 3 issues resolved. 0 new issues introduced. Done in 8.2s. Powered by GLM-5.2.

What AgentGuard does

Not just a scanner — a full pipeline powered by GLM-5.2

🔍

Scan

65 Python rules + 14 JS/TS rules + optional Bandit 100+. Detects injection, secrets, crypto, SSRF, path traversal, prompt injection and more.

🧠

GLM-5.2 Review

Every finding goes to GLM-5.2 for context-aware review. False positives get filtered. Real bugs get confirmed with explanation.

🔧

Auto-Fix

GLM-5.2 generates a fix with full code context. SQL injection to parameterized query. Hardcoded password to env var. eval to ast.literal_eval.

✅

Self-Validate

After fixing, AgentGuard runs ast.parse to verify syntax, re-scans for new issues, and auto-rolls back if anything breaks. GLM-5.2 fallback scan catches anything missed.

⚡

One Command

agentguard pipeline ./src — that is it. Scan, review, fix, validate. Or use individual commands: scan, fix, serve.

🔌

CI/CD Ready

GitHub Action included. SARIF output for GitHub Code Scanning. JSON for custom pipelines. Web GUI on port 1099.

How it compares

Bandit finds bugs. AgentGuard ends them.

BanditSemgrepAgentGuard
Python rules100+Multi-lang65 + Bandit 100+
JS/TS rulesNoYes14
FP filteringNoNoGLM-5.2 Review
Auto-fixNoNoGLM-5.2 Fix
Self-validateNoNoYes
PricingFreeFree/$40Free + Pro $29/mo

Pricing

Start free. Upgrade when you need GLM-5.2 power.

Free

$0
  • 50+ built-in rules + optional Bandit 100+
  • Regex-based auto-fix (offline)
  • Terminal / JSON / SARIF output
  • Web GUI (port 1099)
  • GitHub Action
  • Python + JS/TS scanning
Get Started

Pro

$29/mo
or $199/year (save 43%)
  • 65 Python rules + Bandit 100+
  • GLM-5.2 review (bring own key)
  • GLM-5.2 AI fix (bring own key)
  • Self-validation after fix
  • 100 AI fixes/hour (your own key)
  • Priority support
Try Free 14 Days
Requires your own GLM-5.2 API Key

14-day trial activates automatically — no credit card needed. Requires your own GLM-5.2 API key (free at open.bigmodel.cn). After trial, Free tier continues forever.

Language Support

Python and JavaScript/TypeScript, expanding

Python: 65 rules JavaScript: 14 rules TypeScript: 14 rules + Bandit 100+ (optional)

Detects: eval/exec injection, SQL injection, XSS, hardcoded secrets, weak crypto (MD5/SHA1/ECB), SSRF, SSL issues, path traversal, prototype pollution, pickle deserialization, prompt injection and more.