@syphin/mcp
v0.3.2
Published
Syphin MCP server — centralized AI agent context via Model Context Protocol
Maintainers
Readme
@syphin/mcp
MCP server for Syphin — centralized AI agent context.
Connects Claude Code (or any MCP-compatible agent) to your team's skills, rules, and conventions managed in Syphin. Skills are loaded on demand, cached locally, and usage is tracked via telemetry.
Quick start
Add to Claude Code
claude mcp add syphin -- npx -y @syphin/mcpSet your token
# Option 1: environment variable
export SYPHIN_TOKEN=syphin_your_token_here
# Option 2: credentials file (created by `syphin login`)
# ~/.config/syphin/credentials.jsonInitialize a project
Run syphin init in your repo to create .syphin/config.json, or create it manually:
{
"projectSlug": "your-project",
"environment": "dev"
}How it works
- Claude Code spawns the MCP server via stdio
- The server reads
.syphin/config.jsonfrom your project - It fetches the project manifest from the Syphin API
- Always-on skills are injected into agent context automatically
- Available skills are searchable on demand via MCP tools
MCP tools
| Tool | Description |
|------|-------------|
| search_skills | Search for skills and rules relevant to your current task |
| load_skill | Load the full body of a skill by slug |
| load_rule | Load the full body of a rule by slug |
| list_project_rules | List all skills and rules bound to the current project |
Environment variables
| Variable | Description |
|----------|-------------|
| SYPHIN_TOKEN | Auth token (alternative to credentials file) |
| SYPHIN_API_URL | API base URL (default: https://syphin.vercel.app) |
| SYPHIN_DEBUG | Enable debug logging to stderr |
License
MIT
