Autohand Code

Autohand Code CLI is an open-source, terminal-native AI coding agent built with TypeScript and distributed under the Apache License 2.0. It runs directly in your terminal, reads project context, plans changes, edits files, executes shell commands, and asks for approval before risky operations. The project is hosted at github.com/autohandai/code-cli and has a companion website at autohand.ai/code/.

What It Is

Autohand Code CLI is a command-line AI coding agent designed to keep developers in their terminal without context switching. It combines a reasoning loop with file editing, shell execution, web search, and version control into a single interactive REPL or non-interactive command mode. The agent supports 40+ built-in tools covering file operations, git workflows, planning, memory, browser automation, and multi-agent team coordination.

How the Workflow Works

The agent operates in two primary modes:

• Interactive REPL: Launch autohand for a full session with file mentions (@src/index.ts), slash commands (/model, /skills, /review), skill autocomplete ($frontend-design), and inline terminal commands (! git status).
• Command Mode: Pass --prompt for single-instruction execution suitable for CI/CD pipelines, scripts, or quick tasks. Flags like --yes, --auto-commit, and --dry-run control confirmation and output behavior.

An auto-mode (--auto-mode) runs an autonomous loop with configurable iteration limits, checkpoint intervals, and cost caps, handing off to interactive mode when complete.

Multi-Provider and Model Support

Autohand Code CLI is provider-agnostic. Supported backends include OpenRouter (Claude, GPT-4, Grok), LLMGateway, OpenAI, DeepSeek, Azure Foundry Models, Z.ai, Ollama, llama.cpp, and MLX for Apple Silicon. The active model can be switched mid-session with /model or overridden at launch with --model.

Agent Skills and Extensibility

Skills are modular instruction packages that extend the agent with specialized workflows, functioning like on-demand AGENTS.md files. They can be auto-generated from project analysis (--auto-skill), installed from the community registry at skilled.autohand.ai, or created interactively. The CLI also supports MCP server management (/mcp, /mcp install), custom meta-tools, git hooks, and a Code Agent SDK available in TypeScript, Go, Python, Java, and Swift for embedding the agent runtime in other tools.

Editor and Platform Integration

Beyond the terminal, Autohand Code CLI ships a VS Code extension (AutohandAI.vscode-autohand) available on the VS Code Marketplace and a Zed Editor extension. It runs on macOS, Linux, and Windows. Docker support is documented for containerized workflows. Session checkpointing integrates with Entire (entire.io) for git-backed session rewind and sharing.

Team Workflows and Background Agents

Autohand Code includes a Slack integration so teams can invoke and collaborate with agents from chat, alongside the CLI and editor surfaces. Background agents run autonomously to continue coding, testing, and refactoring across multi-file changes while you're away from the terminal, returning completed work rather than just suggestions.

Update: v0.9.0

The latest release is v0.9.0, published on 2026-05-07, with the repository last pushed on 2026-05-12. The project was created in December 2025 and has accumulated 106 stars and 19 forks on GitHub as of the last recorded update. The roadmap lists enhanced AI model support, a plugin system, advanced test generation, and AI-powered code review as upcoming directions.