# MCP Scripts

> Follow conventions for MCP scripts: CLI args via argparse, call_mcp_tool pattern, and stdout output. Use when writing scripts that wrap MCP tools. Trigger phrases: "write MCP script", "call MCP tool from script", "script conventions".

- Skill: `dojogenesis/mcp-scripts` (Agent Skill)
- Install (CLI): `npx skillmds@latest add dojogenesis/mcp-scripts`
- Raw SKILL.md: https://api.skillmd.com/api/skills/dojogenesis/mcp-scripts/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: DojoGenesis (https://skillmd.com/u/dojogenesis)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/dojogenesis/mcp-scripts

---


# MCP Script Rules

When working with files in `scripts/`:

## DO
- Use CLI arguments for all parameters (argparse)
- Include USAGE docstring at top of file
- Use `call_mcp_tool("server__tool", params)` pattern
- Handle errors gracefully with informative messages
- Print results to stdout for Claude to process

## DON'T
- Hardcode parameters in the script
- Edit scripts to change parameters (use CLI args instead)
- Import from servers/ directly (use runtime.mcp_client)

## Tool Naming
Tool IDs use double underscore: `serverName__toolName`

Examples:
- `morph__warpgrep_codebase_search`
- `ast-grep__ast_grep`
- `perplexity__perplexity_ask`

## Testing
Test with: `uv run python -m runtime.harness scripts/<script>.py --help`