mcp-bridge

Bridge OpenClaw to MCP servers via a local HTTP gateway.

Updated Feb 20, 2026
One-click install
npx skills add https://github.com/kbtime/openclaw-workspace --skill mcp-bridge-kbtime
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-bridge
Source: https://github.com/kbtime/openclaw-workspace/tree/main/skills/mcp-bridge
Command: npx skills add https://github.com/kbtime/openclaw-workspace --skill mcp-bridge-kbtime

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Provides a simple, local HTTP gateway integration that lets OpenClaw discover and invoke tools hosted by Model Context Protocol (MCP) servers without embedding direct server logic into the agent, reducing complexity and centralizing access control.

Core Features & Use Cases

  • Tool Discovery: Query MCP servers for available tools and inspect their JSON input schemas to determine required arguments.
  • Remote Tool Invocation: Execute server-side tools (filesystem reads/writes, vision analysis, web reading, repo searches) and retrieve structured JSON results.
  • Scoped File Access: Enforce whitelist and directory restrictions (e.g., filesystem server limited to /data/workspace) for safer file operations.
  • Use Case: Use the bridge to list fs tools, read a README from a workspace, analyze an image with vision tools, or run webReader to fetch page content.

Quick Start

List tools on the fs server and then call the read_text_file tool to retrieve a file from /data/workspace.

Frequently Asked Questions about mcp-bridge

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I connect OpenClaw to MCP servers for tool discovery and invocation?

To connect OpenClaw to MCP servers, use a local HTTP gateway listening on 127.0.0.1:9712 to bridge requests, discover tool schemas, and execute server-side tool calls programmatically via JSON payloads.

What is an MCP bridge and how does it centralize access control?

An MCP bridge is a local HTTP gateway that lets agents discover and invoke Model Context Protocol tools without embedding direct server logic, centralizing access control through server whitelisting and directory restrictions.

How do I call filesystem MCP tools to read a file from a restricted workspace?

To read a file, query the filesystem MCP server via the local gateway to list available tools, then invoke the read_text_file tool with the target path, enforcing directory restrictions like /data/workspace for safer access.

Can I use the MCP bridge to run vision analysis and web reading tools?

Yes, you can use the MCP bridge to execute server-side tools for vision analysis and web reading, sending JSON requests to the local gateway and retrieving structured JSON results from the respective MCP servers.

Do I need a local gateway running to invoke MCP server tools from OpenClaw?

Yes, invoking MCP server tools from OpenClaw requires a local gateway listening on 127.0.0.1:9712, handling JSON request and response payloads along with basic network and permission error handling.

What are the limitations of using a local HTTP gateway for MCP tool integration?

Limitations of using a local HTTP gateway for MCP tool integration include requiring a local listening port, enforcing strict server whitelisting, and handling basic network and permission errors for scoped file access.