Skip to content
Khadim Khadim Khadim Docs

Khadim

Khadim is an open-source, local-first agentic automation platform. It gives an AI agent the tools to inspect a workspace or screen, write the automation it needs, execute it, recover from failures, and report what happened.

The CLI coding agent is the working entry point today. The desktop app, web control plane, RPA tools, Docker runtime, memory system, and plugins are being built around the same engine.

Highlights

Section titled “Highlights”
  • Run an interactive coding agent in your terminal.
  • Execute one-shot prompts with khadim exec or khadim --prompt.
  • Stream normalized events for apps and services.
  • Use 19+ AI providers through one provider/model interface.
  • Add sandboxed WebAssembly tools through the Plugin SDK.
  • Build toward RPA workflows with screenshots, OCR, browser automation, connectors, schedules, environments, credentials, and memory.
  • Run the same automation model across CLI, desktop, web, Docker, and cloud runners.

Install

Section titled “Install”

Install the CLI from npm:

Terminal window
npm install -g @unravelai/khadim

Start the terminal UI:

Terminal window
khadim

Run one task without opening the UI:

Terminal window
khadim --prompt "explain this codebase"

Learn the basics

Section titled “Learn the basics”

Start with these pages if you are new to Khadim:

  • Installation explains the supported install paths.
  • First steps walks through your first interactive and headless runs.
  • Features maps the current and planned platform capabilities.
  • CLI overview documents the terminal UI and batch modes.

Understand the platform

Section titled “Understand the platform”

Khadim is built around one shared agent loop:

LLM -> plan -> call tools -> observe results -> continue or finish

The same loop powers multiple surfaces:

  • CLI for terminal-first coding and scripted runs.
  • Desktop for local RPA and screen-aware automation.
  • Web for managed agents, team dashboards, and run monitoring.
  • SDK for embedding the agent into applications.
  • Plugins for user-extensible tool domains.

Read Architecture for the full model.

Build with Khadim

Section titled “Build with Khadim”

Use these paths when you want to extend or embed Khadim: