CLI-Anything is HKUDS' (Hong Kong University Data Science) open-source project framed around the thesis 'making ALL software agent-native.' Released in early May 2026, it hit GitHub trending at ★1,047 on May 18 by packaging arbitrary CLI binaries behind a uniform agent-callable interface — letting Claude Code, Codex CLI, OpenClaw, or any other harness invoke previously-unwrapped tooling without bespoke MCP servers per tool. The project ships a registry pattern at clianything.cc and exposes a discoverable hub model where community contributors can register CLI wrappers. The bet: every CLI becomes a skill the moment it has a uniform wrapper, and CLI-Anything is the wrapper convention. Sits adjacent to the skills-as-runtime category but addresses a different layer — the *tool-call surface* rather than the prompt-and-handoff surface.
CLI-Anything addresses a specific gap in the agent-tooling stack: most CLIs were never designed to be called by agents. They expect interactive terminals, ANSI escape codes, prompts for confirmation, and human-readable output. Wrapping each one in a bespoke MCP server is the current standard — and it’s the bottleneck.
The HKUDS approach inverts that. Instead of wrapping each CLI in a dedicated MCP server, CLI-Anything defines a uniform wrapper convention that any CLI can adopt. The result: a single registry of agent-compatible CLIs, callable from any harness that supports the convention.
The skills-as-runtime wave hit a real practitioner ceiling: an agent that can compose skills still needs tools to call. Each new tool requires either an MCP server, a custom integration, or a Python wrapper. CLI-Anything’s bet is that the tool-call surface itself can be standardized — making any CLI agent-callable with a single registration.
For operators running multi-harness stacks (Claude Code + Codex paired, or Cursor + Codex CLI), CLI-Anything is the layer that lets a tool added today work tomorrow across every harness.
★1,047 in the first 24 hours of GitHub trending visibility (May 18, 2026) — day-one signal that the “make CLI tools agent-callable” framing has real practitioner pull. Sits inside the larger 4-day skill-and-agent-tooling category surge that dominated the trending board through mid-May.
Persistent memory layer for AI coding agents — benchmark-backed (95.2% on LongMemEval-S), 92% fewer tokens per session vs full-context pasting, zero manual memory.add() calls.
Open-source AI pair programming tool that works in your terminal to edit code across your entire repository.
AWS's AI-powered coding assistant that helps developers build, deploy, and optimize applications on AWS with code generation and transformation.