The runtimes
that run
skills.

Anthropic's agentic coding tool. Runs in the terminal with full project context, executes commands, edits across files. Skills are a first-class primitive: folders with a SKILL.md that Claude loads when the task matches.

xAI's coding agent. Interactive TUI, headless mode, and ACP support. Fully Claude Code-compatible with zero configuration: it reads Claude Code marketplaces, plugins, skills, MCPs, agents, and CLAUDE.md alongside its own .grok/ directories. Built-in plugin marketplace shipped 2026-06-11.

The self-improving agent from Nous Research. Built-in learning loop: it creates skills from experience, improves them during use, and remembers across sessions. Runs local, Docker, SSH, or serverless (Daytona, Modal), and talks through 20+ messaging platforms.

Open-source, cross-platform personal AI assistant. macOS, Linux, Windows. Extensible through skills and tools, with ClawHub as its public skill registry.

NVIDIA's sandboxed agent runtime. Runs agents in isolated environments with controlled tool access. Solid State's own agent runs on it.

Google's Gemini-powered agent development platform. Spawns parallel agents that browse, type, and verify their own work through artifacts.

OpenAI's open-source coding CLI. Runs locally with sandboxed file and shell access, with bindings for VS Code, Cursor, and Windsurf.

AI-native code editor. Multi-file agent edits, tab autocomplete, background agents, codebase indexing.

Open-source coding agent that runs across terminal, editor, and desktop. Provider-agnostic — works with hosted models and local ones (Ollama included).

Open-source coding agent for VS Code, JetBrains, and Cursor. Plan/Act workflow, checkpoints, built-in MCP support and an MCP marketplace.