Skip to main content

IronClaw Developer Guide

What is this about?

IronClaw is a secure personal AI assistant runtime with a strong emphasis on local control, explicit configuration, and policy-aware execution. This guide explains what IronClaw is, how the newer Reborn runtime is shaped, how setup and configuration work, and where to contribute.

Checked against primary sources

This guide is based on the official IronClaw docs and the official nearai/ironclaw repository, checked on June 26, 2026.

In this section​

1. What IronClaw is​

The shortest accurate description is:

IronClaw is a personal AI runtime designed to be local, explicit, and security-conscious rather than magical and opaque.

The upstream README and docs strongly emphasize:

  • secure personal use,
  • explicit configuration,
  • environment-based secrets,
  • policy-aware execution,
  • and a newer standalone Reborn runtime.

2. The mental model​

Think of IronClaw as five cooperating layers:

LayerWhat it does
CLI runtimeMain user-facing control surface
model routingSelects providers and models explicitly
config and profilesDefine behavior and operating mode
policy and storageControl deployment mode and persistence
skills and automation surfacesExtend assistant capabilities

3. Why developers would choose it​

  • It is deliberately security-focused.
  • It prefers explicit config over hidden defaults.
  • It has a CLI-native configuration flow for providers and models.
  • It is transparent enough that operators can reason about it like a real system instead of a black box.

4. What makes the docs useful​

IronClaw's docs and README do a good job of showing not only how to use it, but also how configuration, policy, secrets, and storage fit together. That is especially valuable for a tool that wants to be trusted.

  1. this overview,
  2. quick start and configuration path,
  3. architecture and runtime,
  4. configuration and security,
  5. contribution surfaces.

Sources​