Skip to content

Guide · Agent skill

How to use the Iron Proxy Gateway Installer agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The add-iron-proxy skill is an agent skill for NanoClaw that installs Iron Proxy, a gateway between a NanoClaw agent and the internet, together with the official Iron Control web console and its database. This guide explains what the skill does, how to install it, how to ask an agent to use it, and what to check before running it. It is written for readers who have not used agent skills before.

Iron Proxy Gateway Installernanocoai/nanoclaw · MIT

Quick start

  1. 1Download and place the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or your project's .claude/skills/ folder. You can also upload it under Settings, Capabilities, Skills in the Claude apps.
  2. 2Check prerequisites. Confirm Docker and pnpm are installed, NanoClaw is running, and the machine can reach the pinned public sources and images.
  3. 3Run setup with the console. From the NanoClaw project root, run the setup script with the --with-control option. Wait for the stage updates to finish.
  4. 4Restart NanoClaw. Restart this NanoClaw copy so it loads the new gateway provider and agent guidance.
  5. 5Ask the agent for a gateway task. Make a specific request, such as adding the Iron console or configuring a credential, and review what the agent reports.

What the add-iron-proxy skill makes the agent do

An agent skill is a folder of instructions and scripts that an AI agent can load when a task matches it. This skill sets up Iron Proxy, a gateway that sits between a NanoClaw agent and the internet. Once installed, outgoing agent requests pass through the proxy.

The proxy checks an allowlist and asks a person to approve requests before they are forwarded. It injects credentials from the Iron Control console, so agents do not see those credentials directly. The skill also adds agent guidance files so the agent knows how to work within this gateway.

What you need before installing

The skill is aimed at developers who already run NanoClaw and are comfortable with Docker. Setup starts several containers and can take several minutes, so it is not a quick plug-in.

  • Docker, which pulls the pinned Iron Control and PostgreSQL images and builds the proxy locally
  • pnpm, which installs the bridge dependencies
  • A running NanoClaw installation
  • Network access to fetch pinned public source and images
  • No GitHub account or API key is required, according to the skill's files

Install the skill from the ZIP or the Claude apps

Download the add-iron-proxy ZIP from this page. Then choose one of two routes. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. Either way, the skill's scripts end up under a path named add-iron-proxy inside the skills folder.

If you use the Claude apps, you can upload the skill instead. Open Settings, go to Capabilities, then Skills, and upload it there.

Run the setup to install Iron Proxy and Iron Control

From the NanoClaw project root, run the setup script with the console option. Setup pulls the pinned Iron Control and database images, starts them on a dedicated Docker network, creates a local operator account, and registers this copy's proxy through Iron's API.

Credentials and encryption keys are stored in owner-only files under data/session-materials/iron-control/. The database has no published port, and neither the console credentials nor the database are mounted into agents. The console port is published only on 127.0.0.1, and the default is 10257. You can change it with NANOCLAW_IRON_CONTROL_PORT in .env before running setup.

  • Setup streams stage names and elapsed-time updates while it runs
  • Source downloads stop after two minutes, the image build after twenty minutes, and console startup after six minutes
  • Setup never opens a Git credential prompt
  • The internal proxy port defaults to 8080 and is set with NANOCLAW_IRON_PROXY_PORT in .env. Re-run setup and restart NanoClaw after changing it
pnpm exec tsx .claude/skills/add-iron-proxy/scripts/setup.ts --with-control

Trigger the skill with natural requests

You do not need to name the skill exactly. The skill's description says it applies when you are selecting the Iron gateway, adding its management UI, configuring credentials and grants, or restoring the proxy, approval bridge, and agent guidance. Plain requests that match those tasks should load it.

Be specific about the goal. Name the gateway, the console, or the credential you want to set up, so the agent can tell this task apart from ordinary coding work.

  • Set up Iron Proxy for NanoClaw and add the Iron Control console.
  • Add the Iron gateway management UI to this NanoClaw install.
  • Configure a credential and grant for the Iron proxy.
  • Restore the Iron proxy, approval bridge, and agent guidance after a change.

Safety notes before you rely on it

This skill changes how the agent handles network traffic and credentials, and it can block agent traffic if it is set up wrongly. Review the installer scripts before running them, as the skill runs scripts and uses the network.

One setting needs particular care. When NANOCLAW_GATEWAY_UNCREDENTIALED_READS is set to true in .env, some GET or HEAD requests skip the approval card. The skill's documentation says this carries a data-leak trade-off, described in docs/gateway-seam.md. Read that file before enabling it.

To remove the gateway, run the removal option of the same setup script. The skill's REMOVE.md says this stops the proxy and console services for this copy, and that NanoClaw's uninstall flow removes this copy's gateway material with its other data.

pnpm exec tsx .claude/skills/add-iron-proxy/scripts/setup.ts --remove

What people use it for

  • Adding human approval before an agent's outgoing web requests are forwarded
  • Storing API credentials in the Iron Control console so agents never receive them directly
  • Restricting agent network access with an allowlist in a local Docker-based gateway
  • Restoring the proxy, approval bridge, and agent guidance after a configuration problem

Frequently asked questions

Is the add-iron-proxy skill free to use?

The skill is published under the MIT licence, as listed on this page. The skill's files do not mention prices, but you still need Docker and the network access setup requires.

What do I need to run the add-iron-proxy skill?

You need Docker, pnpm, a running NanoClaw installation, and network access to fetch the pinned public source and images. The skill's files say no GitHub account or API key is required.

Do I need a GitHub account to install Iron Proxy with this skill?

No. According to the skill's files, the proxy is built locally from the pinned public upstream source, so installation does not require a GitHub account or access to a private proxy image.

Can the agent see my credentials with this skill?

The skill is designed so that credentials are injected by the proxy from Iron Control, and agents do not see them directly. The skill's documentation also says the console credentials and database are not mounted into agents.

Why did setup fail with a timeout or source access error?

Setup prints the last stage it reached. Check which stage failed, such as source download, image build, or console startup, and confirm the machine can reach the pinned source. Fix the reported issue and rerun setup. Keep existing database volumes and encryption keys together.

Does this skill work on Apple Silicon Macs?

The official Iron Control image currently targets linux/amd64. Docker on Apple Silicon runs it with emulation. On other architectures, setup checks the engine first and stops with a command that enables amd64 emulation if the engine cannot run the image.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits NanoClaw developers and teams who want human approval and credential control over agent network access, and who are comfortable with Docker and reviewing installer scripts. If you do not run Docker, want a simple setup, or need a plain writing or productivity skill, pick something else.

Download the skill