Autonomous software engineering agent —
130+ tools · multi-provider LLM · cyberpunk TUI
Built by NICTOLabs · Go + Bubble Tea + LiteLLM
A terminal-based AI coding agent that surpasses all existing solutions. It reads, writes, and executes code autonomously through a visually stunning cyberpunk interface.
Only real, production-ready software. No mocks, no stubs, no placeholders.
No task too complex, no problem too hard. The agent pushes through any challenge.
Work across languages, platforms, and providers. Speed is king.
A battle-tested architecture for autonomous software engineering.
Working, session, persistent, and knowledge base — four tiers of memory for context retention across sessions.
Supervisor patterns, debate protocols, and async swarm execution for complex multi-step tasks.
Permissions, sandboxing, secret scanning — safety rails that don't get in the way of productivity.
State machine, verifier-gated retry, and circuit breaker patterns for robust error recovery.
HITL gateways, progressive delegation, and trust management for safe autonomous operation.
Critique loops, confidence scoring, and trajectory auditing — the agent reflects on its own reasoning.
Prompt optimization, dynamic few-shot selection, and A/B testing across sessions.
MCTS search trees, event triggers, and world models for long-running strategic reasoning.
Thread-safe registry with concurrent execution, parsing, and live streaming.
Switch modes instantly with keyboard shortcuts. Each mode tailors the agent's behavior.
Assist with general coding tasks and questions.
Research and plan without executing any code.
Execute autonomously without asking for permission.
Diagnose and fix issues automatically.
Generate and run comprehensive tests.
Build games with scaffolding and asset generation.
Watch the agent execute real tool calls with animated terminal output — no simulation, just the real thing.
Build from source and start coding with AI.