Hmbown/CodeWhale
An open-source terminal coding agent written in Rust, with MCP support, multi-model routing and multi-agent orchestration.
zinja-coder/zin-mcp-client
ZIN MCP Client is a command-line client for the Model Context Protocol, built as part of the Zin MCP Suite and now archived as read-only by its author. It bridges local language models served by Ollama to STDIO MCP servers, so tools published by those servers can be called by a model that never leaves the machine. The client connects to several MCP servers at once and merges their tools into a single session. Tool selection and invocation are handled by a ReAct agent built on LangChain, which decides which server tool to call for a given request. Three front ends are provided: a formatted interactive CLI, a minimal built-in web UI, and a plug-and-play integration with Open Web UI for users who want a richer chat surface. Detailed logs are written for debugging tool calls and server handshakes. The project targets Python 3.10 or newer and ships as release archives containing zin_mcp_client.py, web_client.py, mcp_proxy.py, an mcp-config.json file and a src directory; dependencies are installed with uv, with pip as a fallback. Its original purpose was testing the MCP servers the same author maintains for mobile reverse engineering, APKTool-MCP-Server and JADX-MCP-Server, and the README also documents use with GhidraMCP for local code review and vulnerability hunting. The author states the project is meant for personal small-scale use with local models and warns that exposing it on a network is a security risk. Development stopped because stronger MCP client alternatives already exist and the initial goal, a lightweight CLI client for testing his own servers, was met. It remains a small readable reference for how a local Ollama model can drive STDIO MCP servers. Licensed under Apache 2.0.
https://github.com/zinja-coder/zin-mcp-client
Also in CLI Tools
yamadashy/repomix
Repomix is a tool that packs entire code repositories into single, AI-friendly files optimized for feeding Large Language Models and other AI tools, with features for token counting, security, and customization.
QwenLM/qwen-code
Open-source terminal coding agent in TypeScript from the Qwen team that consumes MCP servers as tool providers, alongside subagents, skills, hooks, IDE plugins and support for OpenAI, Anthropic, Gemini and Qwen APIs.
1jehuang/jcode
jcode is an extremely RAM-efficient and fast AI coding agent harness for the terminal, designed to optimize multi-session workflows.
txn2/kubefwd
kubefwd is a command-line tool that bulk port forwards Kubernetes services to a local machine, enabling seamless local development by accessing cluster services with their original names and connection strings.
davepoon/buildwithclaude
A comprehensive collection of specialized AI subagents, slash commands, and MCP server integrations designed to enhance the Claude Code AI development environment with powerful automation and domain-specific expertise.
av/harbor
Harbor is a containerized toolkit that enables effortless deployment and management of local large language model backends, frontends, and services with integrated support for the Model Context Protocol ecosystem.
mattzcarey/shippie
Shippie is an AI-powered code review agent that integrates Large Language Models into CI/CD pipelines and functions as a Model Context Protocol client for enhanced integration with external tools.