Installation

Install Ironclad via the interactive installer, from source, or from pre-built binaries. Requires Rust ≥ 1.85 and a C compiler.

Quick Install

Interactive installer (recommended):

bash <(curl -fsSL https://roboticus.ai/install.sh)

Non-interactive (skip all prompts):

curl -fsSL https://roboticus.ai/install.sh | IRONCLAD_YES=1 bash

Installer Steps

Check prerequisites

Verifies a C compiler is available (cc, gcc, or clang)

Install/update Rust

Installs via rustup or updates to stable >= 1.85

Install from crates.io

Runs cargo install ironclad-server --locked (2-5 min)

Verify installation

Checks the binary is on PATH, prints version

Initialize workspace

Creates ~/.ironclad/ with default config, skill templates, empty database

Environment Variables

Variable	Description
`IRONCLAD_VERSION`	Specific version to install (default: latest)
`IRONCLAD_NO_INIT`	Skip 'ironclad init' after install (set to 1)
`IRONCLAD_YES`	Skip all confirmation prompts (set to 1)
`CARGO_HOME`	Custom cargo home (respected if set)

Build from Source

# Clone and build
$ git clone https://github.com/robot-accomplice/ironclad.git
$ cd ironclad
$ cargo build --release

# The binary is at:
# ./target/release/ironclad-server

# Or install to ~/.cargo/bin/
$ cargo install --path crates/ironclad-server

Prerequisites

Rust Toolchain

Rust ≥ 1.85 (edition 2024). Install via rustup.rs

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

C Compiler

Required for bundled SQLite compilation (cc, gcc, or clang).

Linux: sudo apt install build-essential

macOS: xcode-select --install

Post-Install

$ ironclad setup       # Interactive configuration wizard
$ ironclad serve       # Start the agent runtime
$ ironclad dashboard   # Open the web dashboard
$ ironclad --help      # Show all commands

Configuration

Ironclad uses TOML configuration with 14 sections. Minimal config to get started:

[agent]
name = "MyAgent"
id = "my-agent"

[server]
port = 18789

[models]
primary = "ollama/qwen3:8b"

Configuration Sections

Section	Controls
`[models]`	Primary model, fallback chain, routing mode (heuristic or rule), local-first preference
`[providers.*]`	Per-provider URL, tier classification (T1–T4), API keys via env vars
`[memory]`	Token budget allocation across 5 memory tiers
`[cache]`	Exact-match TTL, semantic similarity threshold (default 0.95), max entries
`[treasury]`	Per-payment cap, hourly/daily transfer limits, minimum reserve, daily inference budget
`[skills]`	Skills directory, script timeout, allowed interpreters, sandbox mode, hot-reload
`[a2a]`	Max message size, rate limit per peer, session timeout, on-chain identity requirement