Ironcurtain
A secure* runtime for autonomous AI agents. Policy from plain-English constitutions. (*https://ironcurtain.dev)
What is Ironcurtain?
Ironcurtain is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to secure* runtime for autonomous ai agents. policy from plain-english constitutions. (*https://ironcurtain.dev)
A secure* runtime for autonomous AI agents. Policy from plain-English constitutions. (*https://ironcurtain.dev)
This server falls under the Security category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- A secure* runtime for autonomous AI agents. Policy from plai
Use Cases
Maintainer
Works with
Installation
Manual Installation
npx ironcurtainConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use Ironcurtain
IronCurtain is a secure runtime for autonomous AI agents that enforces behavioral policy written in plain English, independently of model compliance. Rather than trusting the LLM to self-restrict, IronCurtain compiles your plain-English constitution into deterministic policy rules enforced by a separate engine, with agent code running inside a V8 isolate and all external interactions passing through a policy gate. It ships with six pre-configured MCP servers covering filesystem, Git, GitHub, web fetch, Google Workspace, and semantic memory — giving agents broad capabilities while keeping humans in control of mutations through an escalation workflow.
Prerequisites
- Node.js 22 through 25 (Node 22 LTS recommended; versions outside this range are not supported)
- Docker (strongly recommended for Docker Agent Mode which provides additional isolation)
- At least one LLM provider API key: ANTHROPIC_API_KEY, OPENAI_API_KEY, or GOOGLE_GENERATIVE_AI_API_KEY
- An MCP-compatible client or the ability to run IronCurtain directly from the CLI
- Optional: Brave Search, Tavily, or SerpAPI key if you want web search capability through the Fetch MCP server
Install IronCurtain globally
Install the package from npm so the ironcurtain CLI is available system-wide.
npm install -g @provos/ironcurtainRun the guided setup
The setup wizard creates ~/.ironcurtain/config.json, prompts for your LLM API keys, and configures which MCP servers to enable.
ironcurtain setupReview and customize the policy constitution
Edit the plain-English policy file that governs what agents are and are not allowed to do. IronCurtain compiles this into deterministic enforcement rules.
ironcurtain customize-policyCompile the policy
Compile the plain-English constitution into the rule set that the policy engine will enforce at runtime. Re-run this whenever you update the constitution.
ironcurtain compile-policyStart an interactive agent session
Launch the agent runtime with an optional task description. The agent has access to all configured MCP tools and operates under the compiled policy.
ironcurtain start "Refactor the authentication module and open a GitHub PR"Monitor the session in a TUI dashboard
In a separate terminal, use the live dashboard to watch agent progress, review pending escalations requiring human approval, and inspect tool calls.
ironcurtain muxIroncurtain Examples
Client configuration
Configure IronCurtain as an MCP server within Claude Desktop, passing your Anthropic API key.
{
"mcpServers": {
"ironcurtain": {
"command": "npx",
"args": ["ironcurtain", "start", "--stdio"],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}Prompts to try
Example tasks to delegate to an IronCurtain-managed agent.
- "Clone the repo, run the test suite, and fix any failing tests"
- "Search my Gmail for invoices from last month and summarize the totals in a Google Sheet"
- "Find all TODO comments in the codebase and create GitHub issues for each one"
- "Compile the policy and tell me what actions the agent is currently restricted from taking"Troubleshooting Ironcurtain
ironcurtain setup fails because Node.js version is outside the supported range
IronCurtain requires Node.js 22–25. Use nvm to switch versions: `nvm install 22 && nvm use 22`, then re-run setup.
Policy compilation fails with ambiguous rule errors
The compiler rejects policies with conflicting or ambiguous statements. Open the constitution file with `ironcurtain customize-policy` and make each rule more specific — avoid negations of negations and ensure allow/deny scopes do not overlap.
Agent actions are blocked unexpectedly even for benign operations
The default policy is conservative. Run `ironcurtain annotate-tools --all` to see which policy rules apply to each MCP tool, then update the constitution to explicitly allow the operations you need and recompile.
Frequently Asked Questions about Ironcurtain
What is Ironcurtain?
Ironcurtain is a Model Context Protocol (MCP) server that secure* runtime for autonomous ai agents. policy from plain-english constitutions. (*https://ironcurtain.dev) It connects AI assistants to external tools and data sources through a standardized interface.
How do I install Ironcurtain?
Follow the installation instructions on the Ironcurtain GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.
Which AI clients work with Ironcurtain?
Ironcurtain works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is Ironcurtain free to use?
Yes, Ironcurtain is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.
Ironcurtain Alternatives — Similar Security Servers
Looking for alternatives to Ironcurtain? Here are other popular security servers you can use with Claude, Cursor, and VS Code.
Casdoor
★ 13.6kAn open-source Agent-first Identity and Access Management (IAM) /LLM MCP & agent gateway and auth server with web UI supporting OpenClaw, MCP, OAuth, OIDC, SAML, CAS, LDAP, SCIM, WebAuthn, TOTP, MFA, Face ID, Google Workspace, Azure AD
ghidraMCP
★ 9.0kAn Model Context Protocol server that enables LLMs to autonomously reverse engineer applications by exposing Ghidra's decompilation and analysis tools. It allows AI agents to list code structures, rename methods, and analyze binaries directly through
HexStrike AI
★ 8.9kHexStrike AI MCP Agents is an advanced MCP server that lets AI agents (Claude, GPT, Copilot, etc.) autonomously run 150+ cybersecurity tools for automated pentesting, vulnerability discovery, bug bounty automation, and security research. Seamlessly b
IDA Pro MCP
★ 8.7kEnables AI-assisted reverse engineering in IDA Pro by providing tools to analyze binaries, decompile functions, manage comments, search patterns, and interact with the IDA database through natural language.
Anthropic Cybersecurity Skills
★ 6.6k754 structured cybersecurity skills for AI agents · Mapped to 5 frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND & NIST AI RMF · agentskills.io standard · Works with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI & 20+ platform
Hooker
★ 5.1k🔥🔥 hooker is a Frida-based reverse engineering toolkit for Android. It offers a user-friendly CLI, universal scripts, auto hook generation, memory roaming to detect activities/services, one-click SOCKS5 proxy setup, Frida JustTrustMe, and BoringSSL u
Browse More Security MCP Servers
Explore all security servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up Ironcurtain in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use Ironcurtain?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.