🎉 gald3r v5.1.1 is out — Repairing upgrades and review recovery after 5.1.0..See release notes →
🎉 gald3r v5.1.1 is out — Repairing upgrades and review recovery after 5.1.0..See release notes →

OpenHands

FullL2 MCP

All Hands AI

Full CRASH parity — formerly OpenDevin

✅Commands✅Rules✅Skills✅Hooks✅MCPrules: .md

CRASH Capability Matrix

CRASH = Commands · Rules · Agents · Skills · Hooks + MCP. Engine level: L2 — MCP Native

PrimitiveStatus
Commandscommands✅Commands
Rulesrules✅Rules
Skillsskills✅Skills
Hookshooks✅Hooks
MCPmcp✅MCP

Folder Hierarchy

Files gald3r writes and the directories it expects for OpenHands. The .agents/ tree is the primary gald3r surface.

project-root/
.agents/
├── agents/         *.md     ← File-Based Agents
└── skills/         <name>/SKILL.md
.openhands/
└── hooks.json      ← lifecycle hooks
config.toml         ← MCP config
AGENTS.md           ← instruction file

Key Setup Notes

  • →AGENTS.md is the primary instruction file (same as codex/amp)
  • →Skills from .agents/skills/ — shared cross-tool standard
  • →Lifecycle hooks via .openhands/hooks.json
  • →MCP via config.toml

Install gald3r on OpenHands

Run the gald3r installer and get the full CRASH system on OpenHands in under 5 minutes. Persistent memory, 220+ commands, and adversarial quality gates.

install — OpenHands
# 1. Install gald3r: download the installer for your OS from
#    github.com/Gald3r-Labs/gald3r_core/releases (.msi / .pkg / .tar.gz), run it, then confirm
gald3r --version

# 2. Scaffold .gald3r/ in your project
gald3r setup

# 3. Write the OpenHands overlay into this project
gald3r platform install openhands
All platforms
38 platforms supported