@humanity4ai/mcp-servers

MCP action contracts and server runtime for Humanity4AI skills.

All 9 humanity skills are exposed as standard MCP tools using the official @modelcontextprotocol/sdk JSON-RPC 2.0 protocol, natively compatible with Claude Code, Copilot, Manus AI, OpenCode, LangChain, and any other MCP SDK client.

Quick Start for AI Agents

1. Clone and install

git clone https://github.com/humanity4ai/project_human.git
cd project_human
pnpm install

2. Start the MCP server

pnpm start
# or directly:
pnpm --filter @humanity4ai/mcp-servers start

The server starts on stdio using the official MCP SDK JSON-RPC 2.0 protocol. All 9 humanity skills are registered as MCP tools and discoverable via tools/list.

A remote endpoint is also available at https://humanity4ai.ascent.partners/api/mcp using Streamable HTTP transport — no clone needed.

3. Configure your MCP client

Local (stdio):

Add to your MCP client configuration (e.g. claude_desktop_config.json, .cursor/mcp.json, opencode.json):

{
  "mcpServers": {
    "humanity4ai": {
      "command": "pnpm",
      "args": ["--filter", "@humanity4ai/mcp-servers", "start"],
      "cwd": "/path/to/project_human"
    }
  }
}

Or use npx once published to npm (no local clone needed):

{
  "mcpServers": {
    "humanity4ai": {
      "command": "npx",
      "args": ["-y", "@humanity4ai/mcp-servers"]
    }
  }
}

Remote (Streamable HTTP):

{
  "mcpServers": {
    "humanity4ai": {
      "url": "https://humanity4ai.ascent.partners/api/mcp"
    }
  }
}

Available Tools (9 skills)

Protocol

The server uses the official @modelcontextprotocol/sdk and communicates via JSON-RPC 2.0 over two transports:

• stdio — local process (clone & run, npx, Docker)
• Streamable HTTP — remote deployment (Vercel, stateless mode)

Tool discovery:

{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"my-agent","version":"1.0"}}}
{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}

Tool invocation:

{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"empathetic_reframe","arguments":{"message":"I failed the exam I studied so hard for","tone":"warm"}}}

Safety Boundaries

Every tool response includes a boundaryNotice field. Always surface this to users:

• accessibility_audit — Compliance guidance only; does not replace legal review
• rewrite_depression_sensitive_content — Non-clinical UX/content guidance only
• supportive_reply — Non-clinical support; must escalate when risk is elevated (includes grief support modes)
• cognitive_accessibility_audit — Design guidance only
• cultural_context_check — Context-sensitive recommendations with uncertainty disclosure
• deescalation_plan — No coercive tactics
• empathetic_reframe — No manipulation or deceptive empathy
• neurodiversity_design_check — Inclusive design guidance only
• age_inclusive_design_check — Inclusive design guidance only

Package Use

pnpm add @humanity4ai/mcp-servers
# or
npm install @humanity4ai/mcp-servers

Import contracts programmatically:

import { actionContracts, validateContracts } from "@humanity4ai/mcp-servers";
const contracts = validateContracts(actionContracts);

Examples

The examples/ directory contains one complete request/response pair per action:

Roadmap

• Auth policies and rate controls for remote deployments
• Telemetry and evaluation hooks
• Rich scenario scoring beyond structural checks
• MCP Registry listing (mcp-publisher)