trackmcp
Back to directory

Cross-platform C SDK for Model Context Protocol (MCP), in modern🚀 C23.

14 stars CDeveloper Kits Updated Oct 11, 2025
ai-agentai-workflowc23function-callingmcpmcp-c-librarymcp-c-sdkmcp-sdkmcp-servermodel-context-protocolmodern-ctool-calling

Documentation

Table of Contents

Overview

mcpc is a Model Context Protocol library in modern C. Its major goals are:

  • Provide the most native bridge between low-level infrastructure

and AI models

  • Explore the maximum capability of the most recent C programming language

MCP Features

Features

Server

Client

Tool

🔨

Resource

🔨

Prompt

🔨

Completion

🔨

STDIO Transport

🔨

HTTP Transport

🔨

🔨

Others

🔨

🔨

Supported Platforms

 

Linux

Windows

macOS

Linux

Note: this table roughly shows the state of supporting on some platforms,

for more detailed information, check the *Modern C Support*.

Modern C Support

Since mcpc claims to implement everything with the most modern C

possible, its building process becomes a little different than many

existing building systems or process, the major

difference is that: ****we try to support as many C compilers as possible,

even on one platform****.

Ideally mcpc will be implemented fully in C23(currently the newest

standard), but if a full C23 is missing on a platform, we

either downgrade the C standard, or give up the support.

This approach certainly has some benefits and consequences:

1. The library can launch on as many platforms as possible, and be

able to compiled by as many C compilers as possible.

2. There must be a minimum fallback version of C standard that the

library targets, currently it is C11.

3. C23 features that are not patchable, or require non-trivial amount

of effort to patch, will not be used, e.g., N2508, N2645, etc.

4. All safety checks and performance optimization will only focus on the

platforms with full C23 support.

Support Matrix

`✅✅` : *implemented in C23*

`🔨✅` : *implemented in C11~C17*

`-` : *the platform does not exist, or work in progress*

Linux:

 

GLIBC 2.31

GLIBC 2.36

GLIBC 2.40

GCC 10

🔨✅ (Debian 11)

-

-

GCC 12

-

🔨✅ (Debian 12)

-

GCC 13

-

-

-

GCC 14

-

-

✅✅ (Fedora 41)

Clang 14

-

-

-

Clang 16

-

-

-

Clang 18

-

-

-

Clang 19

-

-

✅✅ (Fedora 41)

Windows:

 

CMD/PS

Cygwin

MinGW

MSVC 19

🔨✅

-

-

Clang 19

-

-

-

GCC 14

-

✅✅

✅✅

macOS:

 

macOS 13

macOS 14

macOS 15

Clang 14

🔨✅

-

-

Clang 16

-

🔨✅

🔨✅

GCC 14

🔨✅

-

-

Building

Building (Linux, Cygwin, MSYS2, macOS, etc.)

Prerequisites:

  • GNU Make

Then

cd mcpc

make && make tst && make install

Building (Windows CMD/PS/VS)

Prerequisites:

  • Visual Studio
  • GNU Make (`winget install ezwinports.make`)

Two options to build:

1. Build in CMD/PS

Open "x64 Native Tools Command Prompt", then

cd mcpc

make && make tst

2. Build in Visual Studio:

"File" - "Open" - "CMake", "Build" - "Build All"

> This approach is only for the better debugging experience on Windows,

> *"Build in CMD/PS"* takes priority over this one.

Development

API reference can be found at api.md.

Dependencies:

Code Contributing

Despite we claim to use modern C, as standard as possible, and as

modern as possible, we have some extra conventions, in order to render

our code more readable, inclusive and scalable:

1. Unless mentioned explicitly, we follow GNU C coding style.

2. Prefer functions over macros, use macros only if necessary.

3. The type of variables should be reflected in the source code. Use of

keyword `auto` is thus forbidden.

4. Names of defined variables that are exposed publicly should be in

uppercase, otherwise in lowercase.

FAQ

Should I prefer mcpc over other SDKs?

It depends on your use case. Roughly speaking, if

1. you prefer your MCP servers or clients to perform tasks in the most

native way (e.g. manually manage memory allocation, equipped with

competitive performance, etc.).

2. you wish your MCP servers or clients are distributed with minimum

software requirements (i.e. you don't want to force users to

have Python, Node.js or similar things installed on their machines)

then mcpc is for you. Otherwise I would strongly recommend you consider

Python, Javscript or other SDKs, they have more intuitive and

easy-to-use high-level APIs for MCP protocol.

License

The MIT License (MIT)

Copyright (c) 2025 Michael Lee

Permission is hereby granted, free of charge, to any person obtaining a copy of

this software and associated documentation files (the "Software"), to deal in

the Software without restriction, including without limitation the rights to

use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of

the Software, and to permit persons to whom the Software is furnished to do so,

subject to the following conditions:

The above copyright notice and this permission notice shall be included in all

copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS

FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR

COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER

IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN

CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Frequently asked questions

What is mcpc?

mcpc is Cross-platform C SDK for Model Context Protocol (MCP), in modern🚀 C23.

How do I install mcpc?

Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.

Is mcpc open source?

Yes — it is hosted on GitHub at https://github.com/micl2e2/mcpc and has 14 stars.

Related MCP tools

Run your own MCP server? See who uses it and what to fix.

Measure it with TrackMCP