// Agent profile
Kimi Code CLI is an open-source terminal coding agent from Moonshot AI that reads and edits code, runs shell commands, searches files, and fetches web pages with autonomous planning. It ships as a single binary with no Node.js dependency and launches its TUI in milliseconds. Standout features include subagent dispatching — isolated coder, explore, and plan agents that run in parallel — video input for processing screen recordings as context, conversational MCP server configuration, lifecycle hooks for local commands at decision points, and a plugin ecosystem with marketplace extensions. It integrates with Zed, JetBrains, and other editors via the Agent Client Protocol. Kimi Code is the successor to kimi-cli (10K+ stars) and defaults to Moonshot's Kimi K2.7 Code model, which cuts reasoning token usage by 30% compared to earlier versions.
Kimi Code CLI is Moonshot AI’s terminal-first coding agent, the successor to kimi-cli which accumulated over 10,000 GitHub stars. Where kimi-cli was built on Python, Kimi Code ships as a single binary with a fast Rust-backed TUI that launches in milliseconds. The agent reads and edits code across your repository, executes shell commands, searches the web, and autonomously plans multi-step tasks — adjusting its approach based on feedback from each step. It defaults to Moonshot’s Kimi K2.7 Code model but can be configured to use other compatible providers.
Kimi Code’s most distinctive feature is its subagent system. You can dispatch isolated coder, explore, and plan subagents that run in their own contexts while keeping the main conversation clean. This enables parallel workflows — one subagent refactoring a module while another explores the test suite. Video input is another differentiator: you can feed screen recordings or demo clips directly into the agent as context, useful for reproducing UI bugs or understanding workflows that are hard to describe in text. The conversational MCP configuration lets you add, edit, and authenticate MCP servers through natural language instead of manually editing JSON files. Lifecycle hooks let you attach local commands to critical decision points — for example, running linters before the agent commits code.
Developers use Kimi Code for terminal-native coding workflows: implementing features, debugging, refactoring, and writing tests. The subagent system shines when tackling tasks that naturally decompose — exploring a codebase while simultaneously drafting implementation plans. The video input capability makes it useful for frontend work where a screen recording explains the issue faster than text. Teams that already use MCP servers benefit from the conversational configuration, which lowers the barrier for developers who find JSON configuration files tedious. The plugin ecosystem extends Kimi Code with community-built skills and marketplace extensions.
Kimi Code is MIT-licensed and free to use, but requires API access to a model provider — Moonshot AI’s free tier or a compatible alternative. The project is actively evolving from its kimi-cli predecessor, so the feature set changes quickly. While the subagent system is powerful, it adds complexity compared to simpler single-agent tools. The Kimi K2.7 Code model that ships as the default is optimized for coding but may not match the general reasoning capabilities of larger frontier models on non-coding tasks.
Kimi Code is built for developers who want a fast, terminal-native coding agent with advanced orchestration features. It appeals to developers coming from Claude Code or Aider who want to try a different model ecosystem, teams interested in subagent workflows for parallel task execution, and developers in the Moonshot AI ecosystem who want tight integration with Kimi models. The video input and conversational MCP setup also make it accessible to developers who find existing CLI agents too configuration-heavy.
Builder.io's open-source framework for building agent-native applications — shared actions, SQL-backed state, identity, tools, skills, jobs, observability, and UI surfaces that all work together.
CLI orchestration tool that installs and configures platform connectors for 16+ social and web sources — Twitter/X, Reddit, YouTube, GitHub, Bilibili, and more. Zero API fees.
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.