skills.rest
rcarmo
rcarmoCOMMUNITY
Explore all skills by this author →
Website𝕏@rcarmoGitHub

mcp-adapter

Connects MCP servers to the agent through a unified mcp tool and mcp.json configuration.

847|79|Updated Feb 17, 2026
Search Tags:#tool integration#mcp#Model Context Protocol#server configuration#mcp.json#agent tools#pi-mcp-adapter
One-click install
npx skills add https://github.com/rcarmo/piclaw --skill mcp-adapter
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mcp-adapter
Source: https://github.com/rcarmo/piclaw/tree/main/skel/.pi/skills/mcp-adapter
Command: npx skills add https://github.com/rcarmo/piclaw --skill mcp-adapter

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Configuring and calling Model Context Protocol (MCP) servers inside the Pi agent workspace requires manual wiring; this Skill standardizes how MCP servers are declared in mcp.json and exposed to the agent through one proxied mcp tool.

Core Features & Use Cases

  • Unified MCP proxy: Exposes all configured MCP servers through a single mcp tool with search, describe, and call operations, keeping large tool sets out of the active context.
  • Flexible configuration: Supports project-local config at /workspace/.pi/mcp.json and global config at /config/.pi/agent/mcp.json, with lazy server lifecycle by default.
  • Interactive management: Provides /mcp, /mcp tools, /mcp reconnect, and /mcp-auth commands for inspecting and managing server connections.
  • Use Case: Add a filesystem MCP server to your workspace config, then search for its tools with mcp({ search: "read file" }) and invoke them without restarting your session setup from scratch.

Quick Start

Copy the starter mcp.json example into /workspace/.pi/mcp.json, define your MCP servers, then ask the agent to list available MCP tools using the mcp tool.

Frequently Asked Questions about mcp-adapter

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

FAQPage Schema
How do I configure MCP servers in the Pi workspace?▼

Create a mcp.json file at /workspace/.pi/mcp.json with an mcpServers object defining each server's command and args. You can copy the starter example from /workspace/.pi/mcp.json.example, then restart the session so the adapter reloads the config.

How do I call an MCP tool through the mcp adapter?▼

Call the mcp tool with the tool name and a JSON-string args payload, for example mcp({ tool: "filesystem_read_file", args: "{\"path\":\"./README.md\"}" }). Use mcp({ search: "keyword" }) first to discover the exact tool name.

What is the difference between project-local and global mcp.json?▼

The project-local file at /workspace/.pi/mcp.json ties MCP servers to the current workspace, while the global file at /config/.pi/agent/mcp.json applies across the Pi agent. Prefer the project-local file for workspace-specific servers.

Why are my MCP server changes not taking effect?▼

The adapter reads mcp.json at session start, so edits require restarting the session or reloading Pi. You can also try /mcp reconnect or /mcp reconnect <server> to re-establish individual server connections.

When should I use directTools instead of the mcp proxy?▼

Use directTools only for small, high-value tool sets that you want exposed directly to the agent. Keep large MCP servers behind the mcp proxy to avoid overloading the active tool context.

Related Skills

Explore More Software Engineering Skills→
AradotsoAradotso
7

awesome-mcp-servers-discovery

Discover MCP servers by category with configuration and security guidance.

Official
Basic
HolobiomicsLabHolobiomicsLab
4

protocolsio-integration

Integrate protocols.io data and workflows into MCP deployments.

Official
Advanced
brainbrewlabsbrainbrewlabs
7

mcp-integration

Discover and configure MCP server integrations and test tool connectivity.

Official
Intermediate
kubeworkzkubeworkz

wren-http-api

Initialize sessions, discover tools, and call Wren MCP tools via HTTP JSON-RPC 2.0.

Community
Intermediate