mcps-skill_skill

This skill helps you manage MCP servers and call tools from a CLI, simplifying configuration, daemon control, and tool invocation.
  • Python

2.5k

GitHub Stars

2

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill openclaw/skills --skill mcps-skill

  • _meta.json268 B
  • SKILL.md6.6 KB

Overview

This skill is a CLI manager for MCP (Model Context Protocol) servers that lets you register, run, and call server tools from the terminal. It provides daemon control, server lifecycle commands, and flexible parameter passing for tool invocation. The tool is designed for quick setup, scriptable workflows, and handling mixed parameter types safely.

How this skill works

You register servers in a JSON config (~/.mcps/mcp.json) or via mcps add commands, then run a background daemon to host connections. mcps lists and controls servers, enumerates available tools, and invokes tools with automatic JSON parsing, --raw string mode, or --json file/object input. Environment variables can be used for secrets and runtime options like port and verbose logging.

When to use it

  • When you need a single CLI to manage multiple MCP servers (HTTP, SSE, stdio-backed).
  • When automating tool calls from scripts, CI, or local development with consistent parameters.
  • When you must pass complex JSON, raw SQL, or mixed-type parameters to remote tools.
  • When you need to run a persistent daemon to host and monitor server connections.
  • When you want to archive or back up server configurations and quickly re-create setups.

Best practices

  • Store sensitive credentials as environment variables and reference them in the config using ${VAR_NAME}.
  • Use --raw for parameters that must remain exact strings (IDs, SQL, encoded values).
  • Use --json for large or nested payloads to avoid shell parsing issues and keep commands readable.
  • Keep a single source config in ~/.mcps/mcp.json and use mcps ls and mcps status for quick checks.
  • Start the daemon in verbose mode only when debugging to avoid noisy logs in normal operation.

Example use cases

  • Web scraping: register a fetch stdio server and call fetch url="https://example.com" follow_redirects=true.
  • Database operations: call a Postgres server query with auto-parsed booleans or --raw for SQL strings.
  • Tool discovery: list available tools with mcps tools <server> or mcps tools <server> --simple for names only.
  • Server maintenance: add, disable, restart, or remove servers with mcps add, update --disabled true, restart, and rm.
  • CI integration: script mcps call commands with --json payloads for repeatable automated tasks.

FAQ

Use mcps status to view active connections and mcps ls to inspect configured servers, including disabled ones.

What should I do if a server connection fails?

Start the daemon with mcps start --verbose for detailed logs, then try mcps restart <server> to reconnect.

How do I pass SQL or special-character parameters safely?

Use the --raw flag to keep values as literal strings, or --json to supply a properly escaped JSON payload.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational