Model Context Protocol client hub — connect AI agents to any MCP server for tool discovery, invocation, and ecosystem management. Use when working with MCP clients, discovering MCP servers, or building MCP-based toolchains.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Model Context Protocol client hub — connect AI agents to any MCP server for tool discovery, invocation, and ecosystem management. Use when working with MCP clients, discovering MCP servers, or building MCP-based toolchains.
{"name":"MCP Hub Operator","title":"Model Context Protocol Integrator","expertise":["MCP Client Architecture","Server Discovery & Registration","Tool Invocation","Cross-platform Integration"],"philosophy":"Every agent is only as capable as the tools it can reach.","credentials":["MCP ecosystem maintainer","Tool integration specialist","Cross-platform automation architect"],"principles":["Discover before you build","Standardize the protocol, not the tool","Fail fast with clear error codes","One server, one responsibility"]}
version
1.0.0
MCP Clients Hub — Model Context Protocol Client Ecosystem
When Not to Use
Simple or one-off tasks — if the task is straightforward, direct execution is faster than structured methodology.
Already established workflows — follow existing team conventions rather than introducing new frameworks.
When automation overhead exceeds benefit — for very small scopes, the setup cost may not be justified.
Dependencies
Python 3.8+ or Node.js 18+
Access to relevant APIs/services for your specific use case
Basic understanding of the domain concepts
Commands
# Refer to the skill's usage section for specific commands# Adapt these to your workflow
Money-Making Overview
MCP clients are the gateway layer that connects AI agents to the entire tool ecosystem. Mastering MCP client usage and discovery directly enables revenue generation across every other skill:
Capability
ROI Impact
Timeline
Rapid server discovery & integration
Eliminates 4-8h of manual integration per tool
Day 1
Multi-server orchestration
3-5x more tools per agent = 10x capability surface
Week 1
Custom client for proprietary APIs
Build once, sell as MCP server ($500-5K/server)
Week 2
Tool composition (chaining servers)
Automate 12h workflows in 30 minutes
Week 2
MCP server marketplace listing
Passive income stream ($200-2K/mo per popular server)
Month 1
Total addressable leverage: An agent ecosystem with 50+ connected MCP servers delivers 20x the value of one with 5 servers. Every discovered and integrated server compounds.
# 1. List registered MCP servers and their tools
mcp-discover --list-servers
mcp-discover --server all --list-tools
# 2. Check health of all servers
mcp-discover --health-check --timeout 5s
# 3. Get schema for a specific server's tools
mcp-discover --server github-mcp --schema
Phase 2: Connect and Invoke (15 min)
from mcp_client import MCPClient
# Connect to a discovered server
client = MCPClient("github-mcp", transport="stdio")
# List available tools
tools = client.list_tools()
print(f"Available: {[t.name for t in tools]}")
# Invoke a tool with typed parameters
result = client.call_tool(
"create_issue",
{
"owner": "my-org",
"repo": "my-repo",
"title": "Auto-discovered from MCP client",
"body": "Created via MCP client hub"
}
)
print(f"Status: {result.status}, ID: {result.data.id}")
# Automatically find and connect to new servers as they appearfrom mcp_discover import DiscoveryAgent
agent = DiscoveryAgent(
scan_interval=300, # check every 5 minutes
auto_register=True, # register new servers automatically
health_required=True# only register healthy servers
)
agent.start()
# Handle new servers dynamically@agent.on_server_registereddefhandle_new_server(server_info):
print(f"New server: {server_info.name} ({server_info.transport})")
# Auto-connect and explore tools
client = MCPClient(server_info.name)
tools = client.list_tools()
print(f" Tools: {[t.name for t in tools]}")
# Store in registry for later use
registry.store(server_info)
defsafe_invoke(client, tool_name, params, max_retries=2):
"""Resilient tool invocation with retry and error classification."""
errors = []
for attempt inrange(max_retries + 1):
try:
result = client.call_tool(tool_name, params)
if result.is_error:
error_code = result.error.get("code", -1)
if error_code in (-32700, -32600, -32601): # protocol errorsraise RuntimeError(f"Protocol error: {result.error['message']}")
return result
except ConnectionError as e:
errors.append(f"Connection failed (attempt {attempt + 1}): {e}")
if attempt < max_retries:
time.sleep(2 ** attempt) # exponential backoffcontinueexcept TimeoutError as e:
errors.append(f"Timeout (attempt {attempt + 1}): {e}")
if attempt < max_retries:
time.sleep(1)
continueraise RuntimeError(f"Tool invocation failed after {max_retries} retries: {'; '.join(errors)}")
# Usage
result = safe_invoke(client, "search", {"query": "MCP protocol", "limit": 10})
Step 4: Tool Composition (Chaining)
classMCPPipeline:
"""Compose multiple MCP tools into a reusable pipeline."""def__init__(self, registry: ServerRegistry):
self.registry = registry
self.steps: list[dict] = []
defadd_step(self, server: str, tool: str,
params: dict, transform: callable = None):
"""Add a pipeline step. Transform maps previous output to this step's params."""self.steps.append({
"server": server, "tool": tool,
"params": params, "transform": transform
})
returnselfasyncdefrun(self, initial_context: dict = None):
context = initial_context or {}
results = []
for i, step inenumerate(self.steps):
client = MCPClient(step["server"])
# Resolve params: static dict + transform from previous result
params = dict(step["params"])
if step["transform"] and results:
dynamic = step["transform"](results[-1])
params.update(dynamic)
result = await client.call_tool_async(step["tool"], params)
results.append(result)
context[f"step_{i}"] = result
print(f"[Pipeline] {step['server']}/{step['tool']} → {result.status}")
return context, results
# Example: Research → Document → Notify
pipeline = MCPPipeline(registry)
pipeline.add_step("github-mcp", "search_repos", {"q": "mcp-server", "sort": "stars"})
pipeline.add_step("notion-mcp", "create_database_page", {
"database_id": "research-db",
"properties": {}
}, transform=lambda prev: {
"properties": {
"Name": {"title": [{"text": {"content": f"MCP Server Research - {len(prev.data)} results"}}]},
"Count": {"number": len(prev.data)}
}
})
pipeline.add_step("slack-mcp", "send_message", {
"channel": "#research"
}, transform=lambda prev: {
"text": f"Research documented - see Notion"
})
await pipeline.run({"started_at": datetime.now().isoformat()})
Step 5: Monitoring & Health
defmonitor_servers(servers: list[str], interval_s: int = 60):
"""Continuously monitor MCP server health."""from datetime import datetime, timedelta
status = {s: {"healthy": False, "last_seen": None, "errors": []} for s in servers}
whileTrue:
for name in servers:
try:
client = MCPClient(name)
pong = client.ping()
status[name]["healthy"] = pong
status[name]["last_seen"] = datetime.now()
if pong:
tools = client.list_tools()
status[name]["tools"] = len(tools)
except Exception as e:
status[name]["healthy"] = False
status[name]["errors"].append(str(e))
yield status
time.sleep(interval_s)
# Run health monitorfor snapshot in monitor_servers(["github-mcp", "notion-mcp", "stripe-mcp"], interval_s=30):
unhealthy = [s for s, v in snapshot.items() ifnot v["healthy"]]
if unhealthy:
print(f"WARNING: Unhealthy servers: {unhealthy}")
else:
counts = {s: f"{v.get('tools', 0)} tools"for s, v in snapshot.items()}
print(f"All healthy: {counts}")
Anti-Rationalization Table
Rationalization
Reality
"I'll just use curl to call APIs directly"
MCP handles auth, retries, streaming, type safety, and cross-platform compatibility. Manual curl is tech debt.
"One mega-server is simpler"
Single-responsibility servers are independently deployable, testable, and replaceable. Monoliths rot.
"MCP is just a wrapper around REST"
MCP is a standardized protocol with tool schemas, resource URIs, streaming, and error codes — not an ad-hoc REST convention.
"Discovery is a one-time setup"
Servers come and go, ports change, versions update. Continuous discovery keeps the ecosystem alive.
"I know all the servers I need"
Discovery often surfaces servers you didn't know existed — that is the point.
"Auto-discovery is over-engineering"
Manual registration breaks the moment the ecosystem grows past 5 servers. Automate it.
Output Format
When using MCP clients hub, produce structured results:
