Claude Token Saver MCP runs on your own machine — the client starts it, so there is no endpoint to ping. 501 installs a week from npm. Last commit 31 Mar 2026.
MCP server that offloads routine coding tasks to local LLM (Ollama) to save Claude API tokens
Today is the operative word: we check Claude Token Saver MCP every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add claude-token-saver-mcp -- npx -y claude-token-saver-mcp
{
"mcpServers": {
"claude-token-saver-mcp": {
"command": "npx",
"args": [
"-y",
"claude-token-saver-mcp"
]
}
}
}
[mcp_servers.claude-token-saver-mcp]
command = "npx"
args = ["-y", "claude-token-saver-mcp"]
{
"mcpServers": {
"claude-token-saver-mcp": {
"command": "npx",
"args": [
"-y",
"claude-token-saver-mcp"
]
}
}
}
{
"servers": {
"claude-token-saver-mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"claude-token-saver-mcp"
]
}
}
}
MCP server that reads local Figma .fig files, fully offline
Hand off coding tasks from ChatGPT, Codex or Claude Code to local Oh My Pi (OMP) workers.
MCP server mode for Qoder CLI — delegate coding tasks to local Qoder agents from any MCP client.
Unified MCP server bridging Claude Code and Ollama to Roblox Studio and Blender.
Connect Claude to any OpenAI-compatible LLM endpoint and offload routine work to a local model.
Ready-to-clone Rust MCP server with chat, token-counting, and model tools for Claude
Finds the files a coding task touches, so your AI agent reads less. Local CLI and MCP server.
Local long-term memory MCP server (SQLite) for AI coding agents — OpenCode, Claude Code, Cursor
Answers built from our own checks of this server.