Hermes Skills MCP Server (C)
alex-ht/hermes-skills-mcp-server-c
MCP Server ultracompacto em C para gerenciar skills de agentes IA
Ver no GitHubServidor MCP de alta performance em C puro para descoberta e gerenciamento de skills de IA através do protocolo JSON-RPC 2.0. Otimizado para ambientes de recursos limitados, implementa o padrão SKILL.md com frontmatter YAML e oferece integração nativa com agentes OpenClaw.
Instalação
npx convez-ai alex-ht/hermes-skills-mcp-server-c1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Integração de skills em ambientes edge ou com recursos limitados
- Descoberta automática e catalogação de skills via SKILL.md
- Comunicação JSON-RPC entre agentes IA e repositórios de skills
Veredito

Implementação robusta e minimalista de MCP server para o ecossistema OpenClaw, com foco em performance em C puro. Excelente para deployments com restrições de memória, mas requer compilação e conhecimento técnico para setup.
README
A tradução para o português deste README ainda está na fila.
Agent Skills MCP Server (C Version) ⚡
A high-performance, lightweight, and memory-efficient C implementation of the Agent Skills MCP (Model Context Protocol) Server. This server allows AI agents to discover, view, and manage skills using the standard SKILL.md format with YAML frontmatter.
This implementation is designed for pure OpenClaw environments and is optimized for speed and minimal resource footprint.
🚀 Key Features
- Ultra-Lightweight: Written in pure C, consuming minimal RAM and CPU.
- Zero-Dependency (Mostly): Uses standard
libcjsonandlibyamlfor reliable parsing. - Standard MCP Protocol: Implements the JSON-RPC 2.0 protocol via
stdin/stdoutfor seamless integration with agents. - Workspace Context Aware: Supports the
cwdparameter, allowing agents to switch between different skill directories dynamically. - Automatic Discovery: Scans directories for
SKILL.mdfiles and parses metadata automatically.
🛠️ Prerequisites
To build this server from source, you need a C compiler (like gcc) and the following development libraries:
libcjson-dev(for JSON-RPC communication)libyaml-dev(for parsing skill frontmatter)
On Ubuntu/Debian, you can install them via:
sudo apt-get update && sudo apt-get install -y libcjson-dev libyaml-dev
🔨 Building from Source
-
Clone the repository:
git clone https://github.com/alex-ht/hermes-skills-mcp-server-c.git cd hermes-skills-mcp-server-c -
Compile the project:
make
This will generate an executable named hermes-mcp-server.
🚀 Usage
Running the Server
Once compiled, you can start the server in interactive mode:
./hermes-mcp-server
Integrating with OpenClaw
You can register this server as an MCP tool in OpenClaw by adding it to your configuration:
openclaw mcp set agent-skills '{
"command": "/absolute/path/to/hermes-mcp-server",
"args": []
}'
Example Tool Calls
1. List Skills:
{
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"name": "skills_list",
"arguments": { "cwd": "./skills" }
},
"id": 1
}
2. View a Specific Skill:
{
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"name": "skill_view",
"arguments": { "name": "my-skill", "cwd": "./skills" }
},
"id": 2
}
📁 Project Structure
include/: Header files (skills.h,mcp_utils.h).src/: Source files (main.c,skills.c,mcp_utils.c).Makefile: Build automation script.
📜 License
This project is licensed under the MIT License.