# Kode Agent

> An open-source AI coding assistant that lives in your terminal, capable of understanding codebases, editing files, running commands, and orchestrating multi-model workflows.

Kode is a powerful open-source AI assistant that runs in your terminal and supports multi-model collaboration for complex development workflows. It can understand your codebase, edit files, run shell commands, and delegate tasks to specialized subagents — all from a single CLI. Built with TypeScript and licensed under Apache 2.0, Kode supports 20+ AI models and providers, making it a flexible AI development workbench rather than a single-model assistant.

- **Multi-Model Collaboration** — *Configure unlimited AI model profiles (OpenAI, Anthropic, Alibaba, etc.) and switch between them with `Option+M` or the `/model` command without restarting sessions.*
- **Subagent System** — *Use `@run-agent-name` to delegate tasks to specialized subagents defined in Markdown files; agents can run in parallel for complex workflows.*
- **Expert Model Consultation** — *Use `@ask-model-name` to consult specific AI models for specialized analysis without affecting the main conversation flow.*
- **AGENTS.md Standard Support** — *Natively reads and generates AGENTS.md project documentation files, with legacy `.claude` and `CLAUDE.md` compatibility.*
- **MCP Server Integration** — *Connect to Model Context Protocol servers via `.mcp.json` or `.mcprc` config files to extend tools and context.*
- **Skills & Plugin Marketplace** — *Install agent skills from any git repository using the `add-skill` CLI or manage plugin marketplaces with `kode plugin` commands.*
- **Smart Completion System** — *Fuzzy matching with hyphen-awareness, abbreviation support, and a curated database of 500+ Unix commands for intelligent terminal completions.*
- **Safe & YOLO Modes** — *Runs in YOLO mode by default for maximum productivity; use `kode --safe` to require manual approval for file writes and shell commands.*
- **Cost Tracking** — *Use `/cost` to view per-model token usage and costs across multi-model sessions.*
- **Docker Support** — *Build and run Kode in a container with volume mounts for your project directory and configuration files.*
- **Shareable Model Config** — *Export and import model profiles and pointers as YAML files for team sharing via `kode models export/import`.*
- **ACP Protocol** — *Run Kode as an Agent Client Protocol server (`kode --acp`) for integration with clients like Toad/Zed.*

## Features
- Multi-model collaboration with unlimited configurable AI providers
- Subagent system for task delegation and parallel orchestration
- Expert model consultation via @ask-model-name mentions
- AGENTS.md standard support with legacy .claude compatibility
- MCP server integration for tool and context extension
- Skills and plugin marketplace support
- Smart fuzzy completion with 500+ Unix command database
- Safe mode with manual approval for file writes and shell commands
- Per-model cost tracking and usage statistics
- Docker support
- Shareable YAML model configuration
- ACP (Agent Client Protocol) server mode
- YOLO mode for maximum productivity
- Linux system sandbox via bwrap
- Image paste support on macOS
- Output styles for system-prompt behavior switching

## Integrations
OpenAI, Anthropic Claude, Google Gemini, Alibaba Qwen, Kimi, GLM, Grok, MCP servers, Docker, GitHub, npm, Bun, add-skill CLI, Toad, Zed

## Platforms
WINDOWS, MACOS, LINUX, API, CLI

## Pricing
Open Source

## Links
- Website: https://github.com/shareAI-lab/Kode-Agent
- Documentation: https://github.com/shareAI-lab/Kode-Agent/tree/main/docs
- Repository: https://github.com/shareAI-lab/Kode-Agent
- EveryDev.ai: https://www.everydev.ai/tools/kode-agent