MCP Use Setup Guide¶

Source: mcp-use/mcp-use (10,317 ⭐) Category: Agent Infrastructure / MCP Development

mcp-use is the fullstack MCP framework for developing MCP Apps for ChatGPT, Claude, and MCP Servers for AI agents. It provides Python and TypeScript SDKs with a unified API for building, testing, and deploying MCP-powered applications. Compatible with Hermes, OpenClaw, Claude Code, and any MCP client.

Quick Install¶

# Install the skill
npx skills add mcp-use/mcp-use@mcp-apps-builder

# Python SDK
pip install mcp-use

# TypeScript SDK
npm install mcp-use

Prerequisites¶

Requirement	Details
Python	3.10+ for Python SDK
Node.js	18+ for TypeScript SDK
Hermes Agent	v0.20.0+ with MCP support
MCP Inspector	Optional - for debugging MCP servers

What You Can Build¶

mcp-use unlocks three categories of MCP development:

1. MCP Apps (Agent-facing UI)¶

Build interactive apps that agents can use - dashboards, forms, visualizations. Think of these as "websites for AI agents."

from mcp_use import MCPApp, ui

app = MCPApp("my-dashboard")

@app.page("/")
def dashboard():
    return ui.Card(
        ui.Text("System Status"),
        ui.Chart(data=metrics, type="line")
    )

2. MCP Servers (Tool Providers)¶

Build servers that expose tools, resources, and prompts to MCP clients.

from mcp_use import MCPServer, tool

server = MCPServer("my-tools")

@server.tool()
def search_knowledge_base(query: str) -> list:
    """Search the company knowledge base."""
    return kb.search(query)

3. MCP Clients (Tool Consumers)¶

Build clients that connect to multiple MCP servers and orchestrate tool usage.

Key Features¶

Feature	Description
Unified SDK	Same API across Python and TypeScript
Auto-discovery	Servers auto-register with MCP clients
Hot reload	Edit servers without restarting
Built-in auth	OAuth2.1, API keys, JWT support
Streaming	Real-time tool output streaming
Validation	JSON Schema validation for all tool inputs
Testing	Built-in MCP test harness

Hermes Integration¶

mcp-use servers work natively with Hermes' MCP client. After building a server:

# Start the MCP server
python -m mcp_use serve my_server.py --port 3001

# Register with Hermes (in config.yaml)
mcp_servers:
  my-tools:
    transport: http
    url: http://localhost:3001/mcp

Hermes will auto-discover all tools, resources, and prompts exposed by the server.

Why This Matters¶

Before mcp-use, building MCP servers required manual implementation of the MCP protocol - JSON-RPC message handling, transport layer management, capability negotiation. mcp-use abstracts all of this behind a clean SDK:

Python: pip install mcp-use → @server.tool() decorator → done
TypeScript: npm install mcp-use → server.tool() → done

This reduces MCP server development from days to minutes.

Production Notes¶

10K+ GitHub stars, active daily development
Built by manufact.com - production MCP infrastructure provider
Python package: 100K+ downloads/month (PyPI)
TypeScript package: growing rapidly (npm)
Full documentation at https://mcp-use.com/docs
Active Discord community (https://discord.gg/XkNkSkMz3V)

Troubleshooting¶

Symptom	Likely Cause	Fix
Server not discovered	Wrong transport config	Verify `transport: http` and URL
Tool returns 404	Server not running	Check `mcp_use serve` process
Auth failure	Missing credentials	Configure OAuth or API key in server config

← AgentMemory Setup | Memory Merger Setup → Powered by CorpusIQ

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects