From mcp-openclaw-bridge
Bridge Model Context Protocol (MCP) servers and tools into OpenClaw / ClawHub skills. Publish, discover, and use production MCP capabilities as native agent skills.
How this skill is triggered — by the user, by Claude, or both
Slash command
/mcp-openclaw-bridge:mcp-openclaw-bridgeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Primary use:** When you have (or want to build) an MCP server and need to surface its tools, resources, and prompts as first-class, installable OpenClaw skills that any compatible agent can discover and invoke.
Primary use: When you have (or want to build) an MCP server and need to surface its tools, resources, and prompts as first-class, installable OpenClaw skills that any compatible agent can discover and invoke.
Activate on phrases such as:
A compliant MCP server exposes:
Example production server: the multi-provider MCP server with calculator, web search, file ops, code execution, and dynamic provider routing.
For each logical capability or coherent tool group:
SKILL.md wrapper.clawhub publish.Agent sees the skill, matches trigger phrases, follows the documented mapping, and calls the underlying MCP endpoint (stdio, HTTP, or SDK).
my-mcp-capability/
├── SKILL.md
├── mcp-mapping.json # tool name → MCP method + param transform
├── examples/
│ └── usage.md
└── README.md # human onboarding
Inside SKILL.md include:
MUST declare:
Example declaration block (copy into your bridged skills):
**Data that stays local:**
- All MCP server configuration and credentials
- Any intermediate files or state created by the MCP server
**Data that leaves your machine:**
- Only the serialized tool call arguments for the specific MCP tool invoked
- Standard OpenAI/Anthropic-compatible function call framing (if using remote LLM)
Original MCP tools:
calculator (add, multiply, ...)web_searchfile_operationscode_execution (Python/JS/Bash with sandbox)Bridged OpenClaw skills you can publish:
mcp-calculator (simple arithmetic with verification)mcp-web-research (current information with source grounding)mcp-secure-file-ops (read/write with path validation rules)mcp-code-execution (sandboxed execution + result interpretation)Each becomes its own versioned ClawHub skill with its own trigger table and safety surface.
metadata:
openclaw:
requires:
env:
- MCP_SERVER_PATH
- ANY_REQUIRED_API_KEYS
primaryEnv: MCP_SERVER_PATH
envVars:
- name: MCP_SERVER_PATH
required: true
description: "Path to the MCP server executable or stdio command"
clawhub publish with clear changelog.clawhub install mcp-openclaw-bridge (or the specific bridged skill).Produced by Igor Holt (Genesis Conductor, thermodynamic AI verification, autonomous agent systems) as part of the open agent infrastructure layer. Complements the clawbot-grok persistent state patterns and the production MCP server implementation.
Install via ClawHub (once published):
clawhub install mcp-openclaw-bridge
Use whenever you need to make existing or new MCP servers first-class citizens in the OpenClaw ecosystem.
Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub igor-holt/openclaw-skills --plugin mcp-openclaw-bridge