Agent Inspector Installation Instructions

Two Installation Modes

Agentic IDE Integration: Full secure agent lifecycle from dev to pre-prod. Includes static code scanning, dynamic runtime monitoring, automated correlation between static and dynamic findings, MCP query tools, and AI-powered fixes in Claude Code or Cursor.

Standalone Proxy: Runtime monitoring only with proxy server and dashboard. No IDE dependency.

Agentic IDE Integration Setup

Supports Claude Code and Cursor agentic IDEs

For Claude Code: /plugin marketplace add cylestio/agent-inspector then /plugin install agent-inspector@cylestio then restart

For Cursor: Fetch and follow instructions from https://raw.githubusercontent.com/cylestio/agent-inspector/refs/heads/main/integrations/AGENT_INSPECTOR_SETUP.md then restart

Provides full lifecycle: static analysis, runtime monitoring, correlation, AI-assisted fixes, and pre-prod validation

Standalone Proxy Setup

Install: pipx install agent-inspector (recommended) or uvx agent-inspector openai

Start: agent-inspector openai (or anthropic)

Dashboard at http://localhost:7100, proxy at http://localhost:4000

Runtime monitoring only - no static analysis or IDE integration

Free & Open Source

Choose Your
Installation Mode

Agentic IDE integration for full lifecycle security, or standalone proxy for runtime monitoring only.

Agentic IDE Integration

Full secure agent development lifecycle in Claude Code or Cursor

From static code scanning to runtime validation with automated correlation between static findings and dynamic behavior.

Choose Your Agentic IDE

Currently supporting Claude Code and Cursor

Claude Code

1. Register the Cylestio marketplace:

2. Install the plugin:

Important: Restart Claude Code after installation for the MCP connection to activate.

Cursor

Copy this command to Cursor:

Important: After setup, restart Cursor and approve the MCP server when prompted.

Full Secure Agent Lifecycle

From development to pre-production deployment, complete visibility and control

Static Code Scanning

Pre-runtime scanning for OWASP LLM Top 10 vulnerabilities in your agent code

Runtime Monitoring

Live proxy and dashboard capturing real agent behavior and traces

Static ↔ Dynamic Correlation

Automatically cross-reference code vulnerabilities with actual runtime behavior

MCP Query Tools

Query sessions, risks, and findings directly in your agentic IDE

AI-Powered Fixes

Ask your AI to scan, fix vulnerabilities, and validate changes

Pre-Prod Validation

Validate security posture before deployment with evidence-based reports

Agentic IDE Workflow

Work naturally in Claude Code or Cursor. Ask your AI assistant to scan code, identify issues, generate fixes, run tests, and validate the complete security posture all in one conversational workflow.

Try asking your AI:

"Analyze this agent for security issues"
"Show me runtime validation for these fixes"

Need Help?

Check out the documentation or reach out on GitHub.