Features
Unified meta-harness for Claude Code, Codex, Cursor, Pi, and custom agents
Swap agent harnesses with one-line config changes
Contextual Policies: stateful spend caps, model routing, risk-based escalation
Secure OS sandbox: restrict filesystem/network, broker credentials
YOLO mode with safety controls
Pre-built multi-agent workflows: Polly (coding) and Debby (debate)
Custom agent definitions in YAML
Real-time collaborative sessions shared via URL with full history
Cross-device access: terminal, web, desktop app, mobile, REST API
Credential hiding and access brokering for agents
Command-line interface (CLI) installer via curl, brew, uv, pip
Discord community support
mngr: run any coding agent in parallel anywhere
Bouncer: filter and heal your Twitter feed
Sculptor: manage parallel coding agents with UI
Blueprint: one-shot larger coding tasks
Vet: prevent AI coding agent mistakes
Latchkey: fine-grained HTTP permissions for agents
Ratchets: Rust tool for flexible style violation budgets
Open-source tools and community
Research publications on deep learning and ARC-AGI
Policy work for human-centric AI
Agent behavior auditing and control
Parallel agent orchestration across environments
