ActiveCaveman
🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman
Quick Answer
- 1. Caveman gives AI agents ai & ml capabilities with 90,976 GitHub stars and a quality score of 99/99.
- 2. Install: Run
npx -y cavemanor add the config to your MCP client. - 3. Runtime: npm — no authentication required.
- 4. Health: Actively maintained — last checked 7/29/2026.
Overview
Category: AI & ML
Runtime: npm
Install: npx -y caveman
Auth: No auth required
Source: community
Stars: 90,976
Setup time: ~30 sec
What You Can Build
With the Caveman MCP server, your AI assistant can:
- Access ai & ml functionality through natural language
- Use the server immediately with no API key setup
- Combine with other MCP servers for cross-tool automation workflows
- Run with npm for reliable, consistent deployments
Health Status
ActiveLast commit: 7/20/2026
Checked: 7/29/2026
Page updated: June 13, 2026
One-Click Install
Copy the snippet for your MCP client and paste it in.
Claude Desktop
Add to claude_desktop_config.json
{
"mcpServers": {
"caveman": {
"command": "npx",
"args": [
"-y",
"caveman"
],
"env": {}
}
}
}Cursor
Settings → MCP Servers → Add
{
"mcpServers": {
"caveman": {
"command": "npx -y caveman"
}
}
}Saves as .cursor/mcp.json in the download. Move it to your project root.
VS Code
Use with MCP extension
{
"mcpServers": {
"caveman": {
"command": "npx -y caveman"
}
}
}SDK & Code Integration Examples
Connect to Caveman programmatically using TypeScript or Python MCP SDKs.
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
// Initialize Caveman MCP client transport
const transport = new StdioClientTransport({
command: "npx",
args: ["-y","caveman"],
});
const client = new Client(
{ name: "mcp-app-client", version: "1.0.0" },
{ capabilities: { tools: {} } }
);
async function connectAndRun() {
await client.connect(transport);
const tools = await client.listTools();
console.log("Connected to Caveman. Available tools:", tools);
}
connectAndRun().catch(console.error);MCP Server Configuration
Add this to your claude_desktop_config.json
{
"mcpServers": {
"caveman": {
"command": "npx",
"args": ["-y","caveman"],
"env": {}
}
}
}Topics
Error Resolution & Troubleshooting Matrix
Common JSON-RPC protocol error codes and environment troubleshooting for Caveman.
-32600 (Invalid Request)Cause: Malformed JSON-RPC payload sent to server
Resolution: Verify MCP client payload adheres to JSON-RPC 2.0 specs.
-32601 (Method Not Found)Cause: Requested tool or resource method does not exist
Resolution: Call list_tools() to inspect supported tool names.
-32602 (Invalid Params)Cause: Missing or invalid tool arguments
Resolution: Check argument schema types against tool definition.
-32603 (Internal Error)Cause: Unhandled execution exception inside server process
Resolution: Inspect process stderr logs or check API key credentials.
AUTH_KEY_MISSINGCause: Required environment variable not set in MCP config
Resolution: Define required API key under 'env' object in claude_desktop_config.json.
STDIO_PIPE_CLOSEDCause: Server process crashed on startup
Resolution: Run install command manually in terminal to inspect error output.
Quality Score
Automated quality ranking of maintaining source repository.
Own this server?
Claim your listing to update description, links, and category — and get a verified badge.
Claim this listing →Similar MCP Servers
Other servers in the AI & ML category.
Awesome DeepSeek Integration MCP Server
Model Context Protocol (MCP) server integration for DeepSeek V3 and DeepSeek R1 models. Connect DeepSeek AI reasoning, chat completion, function calling, and code generation to Claude Desktop, Cursor, and AI agents.
DeepSeek MCP Server
MCP server for DeepSeek V3 and R1 models — chat, reasoning mode, code completion, and cost tracking for AI client tools.
Brave Search MCP
MCP server for Brave Search API — web and local search results through AI agents.
OpenAI MCP Server
MCP server for OpenAI — access GPT models, embeddings, and assistants through standard MCP interface.