#8080th of 87 in Coding agents and assistants

kode-cli

Terminal coding agent that edits files and runs commands across many model providers

Stars
5.2k
License
Apache-2.0
Last commit
Aug 2026
Language
TypeScript
Stars, last 29 weeks+631
25 Mar 2026: 4,604 stars1 Apr 2026: 4,749 stars8 Apr 2026: 4,814 stars15 Apr 2026: 4,842 stars22 Apr 2026: 4,881 stars29 Apr 2026: 4,906 stars6 May 2026: 4,930 stars13 May 2026: 4,963 stars20 May 2026: 4,986 stars27 May 2026: 5,016 stars3 Jun 2026: 5,038 stars10 Jun 2026: 5,061 stars17 Jun 2026: 5,081 stars24 Jun 2026: 5,094 stars1 Jul 2026: 5,112 stars8 Jul 2026: 5,119 stars15 Jul 2026: 5,127 stars22 Jul 2026: 5,137 stars29 Jul 2026: 5,152 stars5 Aug 2026: 5,170 stars12 Aug 2026: 5,180 stars19 Aug 2026: 5,187 stars26 Aug 2026: 5,195 stars2 Sept 2026: 5,198 stars9 Sept 2026: 5,204 stars16 Sept 2026: 5,216 stars23 Sept 2026: 5,220 stars30 Sept 2026: 5,226 stars7 Oct 2026: 5,230 stars14 Oct 2026: 5,235 stars

4,604 on 25 Mar 20265,235 on 14 Oct 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Kode is a TypeScript CLI coding agent installed with npm. It reads your repo, edits files and runs shell commands, and routes work across configured models through main, task, compact and quick pointers. It reads AGENTS.md and legacy .claude directories, supports subagents, skills, plugin marketplaces and MCP servers, and can run as an ACP agent server.

Who it is for: Developers wanting a multi-model terminal coding agent

Strengths

  • Model profiles and pointers (main/task/compact/quick) with mid-session switching and YAML export
  • Reads AGENTS.md plus legacy .claude agents, skills and commands
  • No product telemetry by default; network use limited to features you invoke
  • Supports MCP servers, subagents, skill marketplaces and ACP mode for Zed/Toad

Weaknesses

  • Runs in YOLO mode by default, skipping permission checks unless started with --safe
  • Linux bwrap sandbox applies only in safe mode or when enabled by env var
  • No tagged releases on the repository
  • README advises against older Q&A-tuned models such as GPT-4o

What it needs

  • no GPU
  • Docker
  • Needs Node.js (or standalone Bun-compiled binary), ripgrep (bundled via optionalDependencies), bubblewrap (optional, Linux sandbox)
  • Models: Anthropic, OpenAI-compatible endpoints

Also in Coding agents and assistants

See all 87
Also in Coding agents and assistants
RankProjectScore
1claw-codeRust CLI coding agent harness using Anthropic and OpenAI-compatible APIs194.9k starsMIT83 out of 100
2ArchonYAML workflow engine that runs AI coding agents through fixed dev steps23.7k starsMIT82 out of 100
3oh-my-openagentTerminal coding agent with multi-model orchestration and git-backed memory70.1k starscustom license80 out of 100
#4Open SWELangChain coding agent that plans, implements and reviews pull requests11k starsMIT80 out of 100
#5opencodeTerminal coding agent with build and plan modes, plus a desktop app213.8k starsMIT79 out of 100