IronClaw is an open-source, Rust-based autonomous CLI coding agent and execution runtime developed by NEAR AI. It provides an infrastructure harness designed to execute developer workflows, automate code alterations, and manage system operations directly from the terminal. In daily operations, IronClaw works by processing developer prompts against a workspace using codebase memory structures, modular tools, and defined skill registries. The agent invokes language models to inspect repositories, generate patches, and run test suites inside dedicated, platform-aware execution sandboxes to maintain host environment isolation. IronClaw differentiates itself from Python- or TypeScript-based agent runners through its native Rust architecture (structured across modular crates), built-in execution sandboxing, and integrated performance harness for benchmarking agent execution and model latency. It focuses heavily on low runtime overhead and deterministic agent-loop orchestration. The project is open-source under NEAR AI on GitHub and is freely accessible for local builds, modifications, and deployment inside custom runner infrastructures.
Rust-Native Architecture — Implemented as a modular Cargo workspace across dedicated crates for minimal memory footprint and fast native execution. - Platform-Aware Sandboxing — Executes agent commands, terminal scripts, and test runs within isolated sandbox boundaries across supported operating systems. - Codebase Memory Integration — Maintains persistent project state and context across multi-step reasoning cycles using local codebase memory tracking. - Extensible Tool and Skill Registry — Supports custom capabilities and domain-specific operations via organized tool and skill directories. - Latency and Performance Harness — Includes built-in benchmarking utilities to measure inference and execution latency across agent loops. - Container & Runner Deployment — Provides ready-to-use Docker configs and runner scripts for deploying agents in isolated or remote infrastructure.
Use cases
Use Case: Sandboxed Task Execution Developers can deploy IronClaw to execute automated refactoring and build tasks inside sandboxed environments, ensuring that generated shell scripts or test commands cannot inadvertently compromise host machine state.
Use Case: Multi-Step Repository Maintenance Engineers can task the agent with analyzing multi-crate Rust codebases or multi-language projects, using codebase memory to locate dependency mismatches, apply patches, and verify test fixes.
Use Case: Agent Latency and Harness Benchmarking Teams building autonomous agent systems can use IronClaw's latency harness to measure response times, tool-calling overhead, and end-to-end execution bottlenecks across different LLM backends.
Who it is for
IronClaw is designed for systems engineers, AI infrastructure developers, and technical teams looking for a performant, sandboxed Rust runtime to run coding agent loops. It is not intended for non-technical users or developers looking for a simple point-and-click IDE chat plugin.
Install guide
Installation Clone the repository from GitHub and build the binary using Cargo: bash git clone https://github.com/nearai/ironclaw.git cd ironclaw cargo build --release
First Run Set your required LLM provider credentials in your environment and verify the CLI binary: bash export OPENAI_API_KEY="your-api-key" ./target/release/ironclaw --help
Example Session Run a task against your current working directory: bash ./target/release/ironclaw --prompt "Analyze repository structure and run the test harness inside the sandbox."
ActPlane — ActPlane is an open-source agent execution and policy enforcement harness developed by the Eunomia-bpf organization. It functions as a security…
hol-guard — hol-guard is a security scanner and agent guardrail harness developed by Hashgraph Online. Built to safeguard agentic runtime environments and…
Skillreaper — Skillreaper is an open-source skill lifecycle management and auditing tool for AI agent ecosystems, developed by thousandflowers. It analyzes agent…
Agent Island — Agent Island is an open-source agent orchestration harness created by developer tristan666666 on GitHub. Designed as a framework for running…
Stratless — Stratless is an open-source agent harness and execution orchestrator developed by the Stratless AI team. It provides a modular infrastructure…
Google AX — Google AX is an open-source agent harness and execution orchestrator developed by Google. It provides runtime infrastructure and execution sandboxing…