Kode CLI
Kode CLI is a terminal coding agent that reads AGENTS.md and CLAUDE.md, connects MCP servers and works with many models.
- 5,237 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Permission checks are off by default: the agent edits files and runs commands without approval.
- Code and prompts go to the chosen model provider.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/kode-cliDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Permission checks are off by default: the agent edits files and runs commands without approval.
- Code and prompts go to the chosen model provider.
This is third-party code. Review the repository files before installing.
What it does
Kode CLI installs with one npm command and starts as `kode`, `kwa` or `kd`. The agent reads a codebase, edits files and runs terminal commands. Models switch through `/model` across Anthropic- and OpenAI-compatible endpoints, and `@ask-model-name` consults another model. Project instructions come from `AGENTS.md`, `AGENTS.override.md` and `CLAUDE.md`. Extensions connect through MCP (`kode mcp add`, the `.mcp.json` file), `SKILL.md` skills and marketplace plugins. `kode -p` covers non-interactive tasks and the ACP protocol (`kode --acp`) serves clients such as Zed.
Who it is for. Developers who want a Claude Code style terminal agent with model choice and ready compatibility with skills and MCP.
Good fit when
- A project already has `CLAUDE.md` and `.claude`, and another agent should reuse them.
- A task goes to one model while another reviews the architecture through `@ask`.
- A CI script calls `kode -p` to explain a file.
- The Zed editor connects the agent through the ACP protocol.
Not a fit when
- Work involves important files without an isolated environment, since Kode defaults to YOLO mode.
- The model is weak and error-prone on autonomous tasks.
- You only want a quick answer with no edits.
Example request
Read `AGENTS.md`, find why the test in `src/` fails and propose a fix, running commands in `--safe` mode.Limitations
By default Kode runs in YOLO mode: permission checks are off, like `--dangerously-skip-permissions`. Use `kode --safe` for approvals. A key for the chosen model provider is required. On Linux the sandbox needs `bwrap`, or set `KODE_SYSTEM_SANDBOX=0`. Standalone binaries are downloaded separately from the GitHub releases page, and with an npm install using `--no-optional` ripgrep must be installed or pointed to through `KODE_RIPGREP_PATH`.
How to disable. Remove the package with `npm uninstall -g @shareai-lab/kode`. Delete the configuration in `~/.kode.json` and `.kode` directories in projects manually.
FAQ
How do I enable action approvals?
Launch `kode --safe`. Approvals are off by default.
Does Kode read Claude Code files?
Yes. It reads `CLAUDE.md` and `.claude` directories when present, plus `AGENTS.md` and `AGENTS.override.md`.
Does Kode send telemetry?
Not by default. The network is used for model requests, web tools, plugin marketplaces and optional update checks.
How do I install skills?
Run `kode plugin install document-skills@anthropic-agent-skills --scope user` or use `/plugin` inside a session.
Which models are supported?
Anthropic- and OpenAI-compatible endpoints that you configure through `/model` or `kode models import`.
Discussion
No questions or reviews yet.
Related
CLI toolsA terminal coding agent that picks a model for each request on its own and shows spend in real time
A Python terminal coding agent in the manner of Claude Code: reads the repository, edits files, runs commands, with memory, sub-agents and MCP
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
A coding agent with a Python core and three front-ends: a script CLI, a terminal TUI and an Electron app, for OpenAI-style APIs
A Rust coding agent with a native GUI, a terminal mode, an ACP mode for Zed and an MCP server
AI Code Toolkit
aicode-toolkit
Plugins and MCP servers for coding agents: template-based projects, patterns before a file edit and review against YAML rules.
Comparison with similar tools
Cards that share the main search query, and how they differ.
README in short
One npm command installs it, with three command names and compatibility with Codex and Claude Code instructions. MCP, skills and plugins use the same formats.