wevm/incur

CLI framework for agents and humans

What it solves

Incur is a framework for building command-line interfaces (CLIs) specifically optimized for AI agents. It addresses the high token cost and discovery friction associated with traditional MCP (Model Context Protocol) servers or monolithic skill files by providing token-efficient discovery and output formats.

How it works

Incur allows developers to define CLI commands using a lightweight API with Zod schemas for arguments and options. It automates the process of making these tools discoverable to agents through built-in skills add and mcp add commands. To reduce token usage, it uses a custom "TOON" output format that strips unnecessary characters like braces and quotes, and it splits command manifests so agents only load relevant command groups on demand.

Who it’s for

Developers building tools, APIs, or CLIs that need to be seamlessly integrated into AI agent workflows, as well as those wanting to expose their CLI as a standard Fetch API or MCP server over HTTP.

Highlights

  • Agent Discovery: Built-in mechanisms to sync skills and MCP servers so agents can find the CLI automatically.
  • Session Savings: Reduces token consumption by up to 3x compared to traditional MCP or skill alternatives.
  • Call-to-Actions (CTAs): Allows the CLI to suggest the next relevant commands to agents and humans after a run.
  • Flexible Deployment: Can be served as a standalone binary, a standard Fetch API handler, or an MCP server over HTTP.
  • API Integration: Ability to mount existing HTTP servers (like Hono) or OpenAPI specs directly as CLI commands.
  • Type Safety: Uses Zod for well-formed I/O and provides inferred types for command callbacks.

Related

  • Project
    makecindy/cindy

    Cindy is an open‑source, cross‑platform AI‑agent client (Electron desktop + React‑Native mobile) that integrates multiple LLM harnesses (Claude Code, Codex, etc.) with local file access, browser/OS control, persistent memory, and extensible “skills”. The repo contains the client code, shared TypeScript packages, and tooling to build/run the apps; the backend service lives elsewhere. It supports both cloud‑backed usage (via a Cindy account) and fully local agents (skip sign‑in). Licensed under Apache‑2.0.

  • Project
    openclaw/mcporter

    A TypeScript runtime and CLI tool that allows developers and integrate Model Context Protocol (MCP) tools into scripts, terminals, and standalone CLIs.

  • Project
    HKUDS/CLI-Anything

    CLI‑Anything is an open‑source Python framework that automatically generates agent‑friendly command‑line wrappers for any software. It ships a package manager (`cli‑anything‑hub`) that indexes dozens of community‑built CLIs (GIMP, Blender, Zoom, GIS tools, etc.) and lets agents discover, install, and invoke them via structured JSON output. Developers can use a 7‑phase generator to create new CLIs for any application, and AI agents (Claude Code, Cursor, Pi, OpenClaw, …) can install the meta‑skill to browse the registry autonomously. The project is actively maintained, well‑tested (2 461 passing tests), and Apache‑2.0 licensed.

  • Project
    omnimind-ai/OmniInfer

    OmniInfer is a cross‑platform inference engine that lets you run LLMs and vision‑language models locally via a simple CLI and OpenAI‑compatible HTTP server. It supports many backends (llama.cpp, MLX, MNN, etc.), works on desktop, Android and iOS, and provides hardware‑aware optimisation and a terminal UI for chatting.

  • Project
    luoguoxiong/aipack

    A lightweight, extensible TypeScript Agent framework providing a runtime and pluggable ecosystem for building AI agents with built-in support for memory, context compression, and MCP.