Arkya-AI/claude-additional-models-mcp

Reduce Claude Desktop consumption by 10x - Integrate Google's Gemini or Z.ai's GLM-5 (744B params) with Claude via MCP for intelligent task delegation

47 ★8 forksJavaScriptUpdated 6mo ago

What you need to know

MCP server that lets Claude Desktop offload heavy generation and analysis to external models - Google Gemini 3.x (recommended) or GLM-5 (Z.ai) - so Claude focuses on orchestration and file I/O. Exposes 5 tools per integration: ask_/ask__pro for chat and code/reasoning, web_search (grounded), web_reader, and parse_document (PDF/OCR). The repository's CLAUDE.md delegation rules push Claude to spawn parallel sub-agents that call the external model, claiming up to ~96% reduction in Claude token consumption. Gemini uses a hybrid OpenAI-compatible plus native API path; requires GOOGLE_API_KEY or ZAI_API_KEY. Node.js, MIT.

Install

git clone https://github.com/Arkya-AI/claude-additional-models-mcp.git && cd gemini && npm install

Key features

✓5 tools per integration: ask (analysis), ask_pro (code/reasoning), web_search, web_reader, parse_document
✓Gemini 3.x: 1M-token context, 65k max output, native search grounding
✓GLM-5 (Z.ai): 744B parameter model
✓Delegation rules and 3-tier execution priority via CLAUDE.md
✓Parallel sub-agent orchestration pattern

Caveats

⚠Requires an active Google AI or Z.ai API key
⚠Gemini free tier rate limits (15 RPM Flash, 5 RPM Pro)
⚠GLM-5 can interrupt during long tasks

Reviewed 2026-08-11

Topics

ai-toolsanthropicclaude-desktopcost-optimizationglm-5llmmcpmodel-context-protocolproductivityz-ai

Stars: 47★
Forks: 8
Language: JavaScript
License: MIT
Created: 2026-02-14
Last push: 2026-02-15