Automated code-review and fix loop that also produces ML training data. Runs stack-aware review agents, verifies findings against the code, applies fixes validated by the test suite, and records every run as an ATIF trajectory for fine-tuning.
Agent Skills marketplace with 132 active skills for code review, documentation, testing, strategy, research, and planning across Python, Go, Rust, Elixir, React, and iOS. Installs through the skills CLI, Claude Code plugin marketplace, or flat skill directories for Codex-style agents.
Multi-agent orchestrator for software development. Human-in-the-loop approval gates, per-agent model routing, sandboxed execution with credential isolation, real-time dashboard.
Self-improving coding agent in Rust. A directed-search harness rewrites the agent's own prompts and tool descriptions, re-evaluates each candidate end-to-end on a frozen slice of real terminal tasks under a token-penalized reward, and accepts only changes that clear a measured noise margin.