@snippkit/mcp-server
# @snippkit/mcp-server
Official **Model Context Protocol (MCP)** server for **SnippKit**.
Connect your AI coding agents (**Google Antigravity**, **Cursor**, **Claude Desktop**, **Windsurf**, **Claude Code**) directly to your SnippKit library. Empower your agents with procedural **AI Skills**, verified **Code Snippets**, and executable **Shell Commands**.
---
## ⚡ Quick Start
### 1. Cursor Setup
Add the following to your Cursor MCP settings (`.cursor/mcp.json` or Settings > Features > MCP):
```json
{
"mcpServers": {
"snippkit": {
"command": "npx",
"args": ["-y", "@snippkit/mcp-server"],
"env": {
"SNIPPKIT_API_KEY": "sk_live_YOUR_API_KEY"
}
}
}
}
```
### 2. Claude Desktop Setup
Add to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"snippkit": {
"command": "npx",
"args": ["-y", "@snippkit/mcp-server"],
"env": {
"SNIPPKIT_API_KEY": "sk_live_YOUR_API_KEY"
}
}
}
}
```
### 3. Antigravity & Windsurf Setup
In your MCP configuration file:
```json
{
"mcpServers": {
"snippkit": {
"command": "npx",
"args": ["-y", "@snippkit/mcp-server"],
"env": {
"SNIPPKIT_API_KEY": "sk_live_YOUR_API_KEY"
}
}
}
}
```
---
## 🛠️ Provided Tools
| Tool | Description |
| :--- | :--- |
| `snippkit_search_skills` | Search public and team AI Skills in the SnippKit registry by keywords or agent compatibility. |
| `snippkit_get_skill` | Fetch complete `SKILL.md` instructions and procedural rules to guide agent execution. |
| `snippkit_search_snippets` | Search your private and public code snippets by title, tag, or language. |
| `snippkit_get_snippet` | Retrieve decrypted code from your SnippKit library. |
| `snippkit_save_snippet` | Save newly written or refactored code directly into SnippKit from the agent. |
| `snippkit_search_commands` | Search verified shell scripts and DevOps command runbooks. |
| `snippkit_get_command` | Retrieve shell command scripts. |
---
## 📚 Resources & Prompts
- **Resources:**
- `snippkit://skills/{slug}`: Read a skill's full instructions as a direct context resource.
- `snippkit://snippets/{id}`: Read a snippet as a direct context resource.
- **Prompts:**
- `use-skill`: Injects a named SnippKit skill into the agent's context window.
- `save-solution`: Prompts the agent to format and persist code to SnippKit.
---
## 🔐 Authentication
Generate your API key from [SnippKit Dashboard > Settings > Developer Keys](https://snippkit.com/settings/developer).
The server checks credentials in this order:
1. `SNIPPKIT_API_KEY` environment variable.
2. Command-line flag `--api-key <key>`.
3. Local `snix` CLI vault (if you have already logged in via `snix login`).
---
## 📄 License
MIT © [SnippKit](https://snippkit.com)
TDQS
Scored across 7 tools
Each tool pairs a clear verb (search/get/save) with a distinct resource (skills, snippets, commands), so overlaps are minimal. The only mild ambiguity is conceptual: skills, snippets, and commands can all contain code-like content, so an agent might briefly hesitate between search_skills and search_snippets, but the descriptions resolve it.
All names follow a uniform `snippkit_<verb>_<resource>` snake_case pattern, which is highly predictable. The only deviation is singular vs. plural resource nouns (search_skills vs get_skill), a minor inconsistency that doesn't impede selection.
Seven tools is well-scoped for a registry/library server, with a symmetric search+get pair per resource and a single write operation. Every tool earns its place and nothing feels redundant or padded.
Search and retrieval are covered symmetrically across skills, snippets, and commands, and snippets support creation. However, snippets lack update/delete operations and the write path is asymmetric (no save for commands, no delete anywhere), leaving notable lifecycle gaps.