1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361# Build server: https://github.com/modelcontextprotocol/servers/tree/main/src/brave-search#build
from __future__ import annotations
import json
import logging
import os
from collections.abc import Iterable, Mapping
from contextlib import AsyncExitStack
from dataclasses import dataclass
from typing import Any, Literal, NotRequired, Protocol, TypedDict, get_args
import colorama
from anthropic import Anthropic
from anthropic.types import ContentBlock, ToolUseBlock
from dotenv import load_dotenv
from mcp import ClientSession, StdioServerParameters, Tool, stdio_client
from mcp.client.stdio import get_default_environment
from openai import OpenAI
from openai.types.chat import (
ChatCompletionMessage,
ChatCompletionMessageToolCall,
ChatCompletionToolParam,
)
load_dotenv()
colorama.init(autoreset=True)
MAX_TOKENS = 1000
logger = logging.getLogger("my_mcp_client")
SupportedModels = Literal[
"gpt-4o-mini", "gpt-4o", "gemini-2.0-flash", "claude-3-5-sonnet-latest"
]
ServerName = str
ToolName = str
class MCPServerParameter(TypedDict):
command: str
args: list[str]
env: NotRequired[list[str]]
class MCPClient:
def __init__(self, llm_api_client):
self.exit_stack = AsyncExitStack()
self.llm_api_client = llm_api_client
self.available_tools: list[Tool] = []
self.tool2session: dict[ToolName, ClientSession] = {}
async def connect_to_server(self, servers: Mapping[ServerName, MCPServerParameter]):
sessions: dict[ServerName, ClientSession] = {}
for name, parameters in servers.items():
env = get_default_environment()
for env_name in parameters.get("env", []):
env[env_name] = os.getenv(env_name)
server_params = StdioServerParameters(
command=parameters["command"],
args=parameters["args"],
env=env,
)
stdio_transport = await self.exit_stack.enter_async_context(
stdio_client(server_params)
)
sessions[name] = await self.exit_stack.enter_async_context(
ClientSession(*stdio_transport)
)
for name, session in sessions.items():
await session.initialize()
response = await session.list_tools()
self.available_tools.extend(response.tools)
for tool in response.tools:
self.tool2session[tool.name] = session
logger.debug(
"Session of server %s is initialized and tools are listed", name
)
logger.info(
"Connection to server with tools: %s",
[tool.name for tool in self.available_tools],
)
def completion(self, messages):
return self.llm_api_client.completion(
messages,
max_tokens=MAX_TOKENS,
tools=self.llm_api_client.format_tools(self.available_tools),
temperature=0.0,
)
def get_text(self, message):
return self.llm_api_client.get_text(message)
def get_tool_calls(self, message):
return self.llm_api_client.get_tool_calls(message)
async def call_tools(self, tool_calls):
return [await self.call_tool(tool_call) for tool_call in tool_calls]
async def call_tool(self, tool_call):
tool = self.format_tool_call(tool_call)
tool_result = await self.tool2session[tool.name].call_tool(tool.name, tool.args)
logger.info(
"[Calling tool %s with args %s, Got %s]", tool.name, tool.args, tool_result
)
return self.llm_api_client.format_tool_result(tool_result.content, tool)
def format_tool_call(self, tool_call):
return self.llm_api_client.format_tool_call(tool_call)
async def cleanup(self):
await self.exit_stack.aclose()
class ChatSession:
def __init__(self, client: MCPClient) -> None:
self.mcp_client = client
self.messages = []
def print_empty_line(self):
print()
async def start(self) -> None:
logger.info("MCP Client Started!")
print("Type your queries or `quit` to exit.")
while True:
try:
self.print_empty_line()
query = input("Query: ").strip()
if query == "":
continue
if query.lower() == "quit":
break
response = await self.process_query(query)
self.print_empty_line()
print(response)
except Exception as e:
logger.exception("Error: %r", e)
self.print_empty_line()
print("γ¨γ©γΌγηΊηγγΎγγγη΅δΊγγΎγ")
break
async def process_query(self, query: str) -> str:
self.messages.append({"role": "user", "content": query})
message = self.mcp_client.completion(self.messages)
self.messages.append(message)
while tool_calls := self.mcp_client.get_tool_calls(message):
for tool_call in tool_calls:
call = self.mcp_client.format_tool_call(tool_call)
print(f"{colorama.Fore.CYAN}Call tool {call.name!r} with {call.args}")
tool_call_messages = await self.mcp_client.call_tools(tool_calls)
for tool_call_message in tool_call_messages:
print(f"{colorama.Back.CYAN}{tool_call_message["content"]}")
print()
self.messages.extend(tool_call_messages)
message = self.mcp_client.completion(self.messages)
self.messages.append(message)
return self.mcp_client.get_text(message)
async def main(llm_api_client: LlmWebApiClient) -> None:
with open("servers.json") as f:
servers = json.load(f)
mcp_client = MCPClient(llm_api_client)
try:
await mcp_client.connect_to_server(servers["mcpServers"])
chat_session = ChatSession(mcp_client)
await chat_session.start()
finally:
await mcp_client.cleanup()
@dataclass
class GenericToolCall:
id: str
name: str
args: dict[str, Any]
class LlmWebApiClientInterface(Protocol):
def format_tools(self, tools: Iterable[Tool]): ...
def completion(self, messages, max_tokens, tools, temperature): ...
def get_text(self, message): ...
def get_tool_calls(self, message): ...
def format_tool_call(self, tool_call) -> GenericToolCall: ...
def format_tool_result(self, result_content, tool_call: GenericToolCall) -> str: ...
class OpenAICompatibleApiClient:
client: OpenAI
model_name: str
def format_tools(self, tools: Iterable[Tool]) -> list[ChatCompletionToolParam]:
return [
{
"type": "function",
"function": {
"name": tool.name,
"description": tool.description,
"parameters": tool.inputSchema,
},
}
for tool in tools
]
def completion(
self, messages: list[ChatCompletionMessage], max_tokens: int, tools, temperature
) -> ChatCompletionMessage:
logger.debug(messages)
response = self.client.chat.completions.create(
model=self.model_name,
max_tokens=max_tokens,
messages=messages,
tools=tools,
temperature=temperature,
)
logger.debug(response)
assert len(response.choices) == 1
return response.choices[0].message
def get_text(self, message: ChatCompletionMessage) -> str:
return message.content
def get_tool_calls(self, message: ChatCompletionMessage) -> bool:
return message.tool_calls
def format_tool_call(
self, tool_call: ChatCompletionMessageToolCall
) -> GenericToolCall:
return GenericToolCall(
id=tool_call.id,
name=tool_call.function.name,
args=json.loads(tool_call.function.arguments),
)
def format_tool_result(self, content, tool_call: GenericToolCall):
tool_result_contents = [content.model_dump() for content in content]
return {
"role": "tool",
"tool_call_id": tool_call.id,
"name": tool_call.name,
"content": tool_result_contents,
}
class GPTApiClient(OpenAICompatibleApiClient):
def __init__(self, model_name: str) -> None:
self.client = OpenAI()
self.model_name = model_name
class GeminiCompatibleOpenAIApiClient(OpenAICompatibleApiClient):
def __init__(self, model_name: str) -> None:
self.client = OpenAI(
api_key=os.getenv("GOOGLE_API_KEY"),
base_url="https://generativelanguage.googleapis.com/v1beta/",
)
self.model_name = model_name
class ClaudeApiClient:
def __init__(self, model_name: str) -> None:
self.client = Anthropic()
self.model_name = model_name
def format_tools(self, tools: Iterable[Tool]):
return [
{
"name": tool.name,
"description": tool.description,
"input_schema": tool.inputSchema,
}
for tool in tools
]
def completion(
self, messages, max_tokens, tools, temperature
) -> list[ContentBlock]:
logger.debug(messages)
response = self.client.messages.create(
model=self.model_name,
max_tokens=max_tokens,
messages=messages,
tools=tools,
temperature=temperature,
)
logger.debug(response)
return {"role": "assistant", "content": response.content}
def get_text(self, message) -> str:
content = message["content"]
assert len(content) == 1
return content[0].text
def get_tool_calls(self, message) -> list[ToolUseBlock]:
contents: list[ContentBlock] = message["content"]
return [content for content in contents if content.type == "tool_use"]
def format_tool_call(self, tool_call: ToolUseBlock) -> GenericToolCall:
return GenericToolCall(
id=tool_call.id,
name=tool_call.name,
args=tool_call.input,
)
def format_tool_result(self, content, tool_call: GenericToolCall) -> str:
return {
"role": "user",
"content": [
{"type": "tool_result", "tool_use_id": tool_call.id, "content": content}
],
}
def LlmWebApiClient(model_name: str) -> OpenAI:
if model_name.startswith("gpt"):
return GPTApiClient(model_name)
elif model_name.startswith("gemini"):
return GeminiCompatibleOpenAIApiClient(model_name)
elif model_name.startswith("claude"):
return ClaudeApiClient(model_name)
raise ValueError(f"Unsupported model name: {model_name}")
if __name__ == "__main__":
import argparse
import asyncio
parser = argparse.ArgumentParser()
parser.add_argument("model_name", choices=get_args(SupportedModels))
parser.add_argument("-v", "--verbose", action="store_true")
args = parser.parse_args()
my_mcp_client_logger = logging.getLogger("my_mcp_client")
my_mcp_client_logger.setLevel(logging.DEBUG if args.verbose else logging.INFO)
my_mcp_client_handler = logging.FileHandler("mcp-client.log")
my_mcp_client_handler.setFormatter(
logging.Formatter(
"%(asctime)s | %(levelname)s | %(name)s:%(funcName)s:%(lineno)d - %(message)s"
)
)
my_mcp_client_logger.addHandler(my_mcp_client_handler)
client = LlmWebApiClient(args.model_name)
asyncio.run(main(client))