Claude Desktop Integration

Use Inkog directly within Claude Desktop for natural language security analysis of your AI agents.

Installation

~/Library/Application Support/Claude/claude_desktop_config.json

%APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "inkog": {
      "command": "npx",
      "args": ["-y", "@inkog-io/mcp"],
      "env": {
        "INKOG_API_KEY": "sk_live_your_api_key"
      }
    }
  }
}

Verify Installation

Ask Claude:

“What Inkog tools do you have access to?”

Claude should list the 7 available tools.

Example Workflows

Security Scan

“Use Inkog to scan my LangChain agent at ~/projects/my-agent”

Governance Check

“Verify the AGENTS.md governance file in my project”

Compliance Report

“Generate an EU AI Act compliance report for my agent”

Remediation

“Explain how to fix prompt injection vulnerabilities”

Agent-Assisted Development Workflow

With Inkog in Claude Desktop, you can use a security-first development workflow:

1. Build — Ask Claude to help you build your AI agent
2. Scan — “Use Inkog to scan my agent for security issues”
3. Fix — “Explain and fix the prompt injection finding”
4. Govern — “Create an AGENTS.md and verify it matches the code”
5. Comply — “Generate an EU AI Act compliance report”
6. Ship — “Add Inkog to my GitHub Actions CI pipeline”

Claude will use the appropriate Inkog tools at each step, providing contextual security guidance alongside code generation.

Troubleshooting

Next Steps

• MCP Server Reference - Full tool documentation
• AGENTS.md Governance - Learn about governance verification