Skip to main content
Glama

Humanity4AI

9 humanity skills for AI agents โ€” crisis detection, accessibility auditing, empathy, cultural sensitivity, and more. Ready-to-use via MCP, npm, or direct prompting.

CI npm License: MIT Release CodeQL MCP Registry Works with

Demo


Skills at a Glance

Skill

Category

What it does

๐Ÿ›ก๏ธ Supportive Reply

Emotional Safety

Detects crisis signals, generates supportive responses with escalation guidance

๐Ÿ“ Safe Content Rewriter

Emotional Safety

Audits and rewrites text to remove harmful, stigmatising, or triggering patterns

โ™ฟ Accessibility Audit

Accessibility

Scores pages against all 86 WCAG 2.2 success criteria (A/AA/AAA)

๐Ÿง  Cognitive Accessibility

Cognitive Support

Audits content for reading level, structure, and cognitive load

๐ŸŒ Cultural Context Check

Cultural Context

Flags cultural sensitivity issues for a given audience and region

๐Ÿ”ฅ De-escalation Plan

Conflict Navigation

Generates structured de-escalation plans calibrated to conflict intensity

๐Ÿ’ฌ Empathetic Reframe

Communication

Reframes messages with genuine empathy, catching hollow empathy patterns

๐Ÿงฉ Neurodiversity Design

Neurodiversity

Audits UIs for ADHD, autism, dyslexia, and sensory sensitivity

๐Ÿ‘ถ Age-Inclusive Design

Age Inclusion

Audits user flows for age barriers across children, adults, and older users


Related MCP server: mcp-llm

Quick Start

Zero setup โ€” just a URL:

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

Or one command with npx:

npx @humanity4ai/mcp-servers

Then configure your MCP client:

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

Docker: docker compose up
For contributors: git clone โ€ฆ && pnpm install && pnpm start

All 9 skills are discoverable via tools/list and invocable via tools/call.

Prerequisites (for local only): Node.js >= 22, pnpm >= 10 | Windows, macOS, Linux, Android, iOS


Architecture

graph TD
    A[MCP Client<br/>VS Code / Cursor / Claude Code] -->|stdio| B[mcp-server.ts]
    A -->|HTTP POST| C[Vercel<br/>humanity4ai.ascent.partners]
    B --> D[server-factory.ts<br/>createServer]
    C -->|api/mcp.ts| D
    D --> E[9 Tool Registrations]
    E --> F[handlers.ts<br/>invokeAction]
    F --> G[validate.ts<br/>SCHEMA_REGISTRY]
    G --> H[schemas-data.ts<br/>18 inline schemas]
    F --> I[9 Handler Functions<br/>rule-based, zero LLM calls]
    I --> J[crisis-resources<br/>patterns<br/>i18n<br/>wcag-criteria]

Four Ways to Use

Method

Best for

How

Remote URL

Zero-setup, any MCP client

Point to https://humanity4ai.ascent.partners/api/mcp

MCP Server

VS Code, Cursor, Claude Code, Copilot, Manus AI, OpenCode

npx @humanity4ai/mcp-servers or clone & start

LLM Prompting

ChatGPT, Claude, Gemini (web chat)

Share llms.txt or paste llms-full.txt into the chat

Local Files

Offline CLI tools

Clone the repo, point your tool at the skills/ directory

See Agent Adapter Guide for platform-specific setup instructions.


Why Humanity4AI?

AI agents are everywhere โ€” but they're not always humane. Humanity4AI gives agents reusable, tested skills for the moments that matter:

  • An agent detects a user in crisis โ†’ Supportive Reply generates an appropriate response and escalates to qualified help

  • A UI is inaccessible to screen readers โ†’ Accessibility Audit scores it against WCAG 2.2 and provides remediation

  • Content uses stigmatising language โ†’ Safe Content Rewriter flags and rewrites harmful patterns

  • A conflict is escalating โ†’ De-escalation Plan generates structured, non-coercive guidance

Every skill includes:

  • Explicit safety boundaries โ€” what the skill can and cannot do

  • Uncertainty disclosure โ€” confidence level stated upfront (low/medium/high)

  • Evaluation gates โ€” automated baseline checks for quality

These skills are rule-based and non-clinical โ€” they do not provide diagnosis, treatment, or professional medical/legal advice.


Supported Platforms

OpenCode ยท Claude Code ยท Microsoft Copilot ยท Manus AI ยท OpenClaw ยท ChatGPT ยท Claude ยท Gemini


Contributing

git clone https://github.com/<your-username>/project_human.git
cd project_human
git checkout -b my-contribution
# Copy the skill template if adding a new skill:
cp -r templates/skill skills/my-skill-name
pnpm check && pnpm evals && pnpm test  # Run all checks before PR

Open a PR targeting main. Browse good first issues or read the Contributing Guide.

Error

Fix

ERR_PNPM_OUTDATED_LOCKFILE

Run pnpm install, then commit pnpm-lock.yaml

pnpm: command not found

npm install -g pnpm

pnpm evals fails

EVAL_REPORT=1 pnpm evals โ€” see evals/reports/latest.md


Resources

License

MIT โ€” see LICENSE.


A Humanity4AI project by Ascent Partners Foundation.

If this saves you time, a โญ on GitHub helps others find it.

Available Tools

9 tools
accessibility_auditA

Audit web pages for WCAG 2.2 compliance or set session WCAG level. Crawl mode: provide pages with URL+HTML for per-page scoring and ranking across all 86 success criteria. Session mode: returns the complete WCAG checklist for the requested level that the agent must enforce.

Inputs: url, mode, urls, level, +2 more.

age_inclusive_design_checkA

Audit a user flow or interface for age-inclusive design โ€” covering children, adults, and older users. Returns recommendations to remove age-related barriers. Safety boundary: inclusive design guidance only.

Inputs: flow_description (required), age_groups.

cognitive_accessibility_auditA

Audit content for cognitive accessibility โ€” plain language, reading level, structure, and clarity. Returns actionable recommendations to reduce cognitive load. Safety boundary: design guidance only.

Inputs: content (required), target_context.

cultural_context_checkA

Check a message for cultural sensitivity issues for a given audience or region. Returns flags, concerns, and suggested alternatives with uncertainty disclosure. Safety boundary: context-sensitive recommendations with uncertainty disclosure.

Inputs: message (required), audience (required), region.

deescalation_planA

Generate a structured de-escalation plan for a conflict or tense situation. Returns step-by-step guidance calibrated to the intensity of the conflict. Safety boundary: no coercive tactics.

Inputs: situation (required), intensity.

empathetic_reframeA

Reframe a message with genuine empathy โ€” acknowledging emotions, validating experience, and offering presence. Detects and replaces hollow empathy phrases with authentic alternatives. Safety boundary: no manipulation or deceptive empathy.

Inputs: message (required), tone.

neurodiversity_design_checkA

Audit a UI description for neurodiversity-aware design โ€” covering ADHD, autism, dyslexia, and sensory sensitivities. Returns targeted recommendations for inclusive design. Safety boundary: inclusive design guidance only.

Inputs: ui_description (required), focus.

rewrite_depression_sensitive_contentA

Audit or rewrite text to be sensitive to depression and mental health. Removes harmful language patterns and replaces them with supportive, non-stigmatising alternatives. Safety boundary: non-clinical UX/content guidance only.

Inputs: text (required), mode, domain.

supportive_replyA

Generate a supportive, non-clinical reply to a message from someone in distress. Supports general emotional support and grief-specific modes (presence, practical, reflection). Includes escalation guidance calibrated to the assessed risk level. Safety boundary: non-clinical support only; must escalate when risk is elevated.

Inputs: message (required), risk_level (required), locale, support_mode.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelogโ€ฆ

    TDQS

    A4.1/5.0

    Scored across 9 tools

    Disambiguation5/5

    Each tool has a distinct purpose: WCAG accessibility audits, depression-sensitive rewriting, supportive replies, cognitive accessibility, cultural context checks, de-escalation plans, empathetic reframing, neurodiversity design, and age-inclusive design. No overlap; agents can easily differentiate.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern in snake_case (e.g., accessibility_audit, rewrite_depression_sensitive_content, cultural_context_check). Predictable and uniform.

    Tool Count5/5

    With 9 tools, the set is well-scoped for the domain of human-centered design and sensitive communication. Each tool earns its place without being too few or too many.

    Completeness5/5

    The tools comprehensively cover accessibility (WCAG, cognitive, neurodiversity, age), mental health sensitivity, cultural sensitivity, and communication guidance. There are no obvious gaps for the stated purpose of inclusive and sensitive interaction.

    Maintenance

    ActivitySlowing
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      A MCP server enabling LLMs to roll dice
      1
      4
      MIT
    • A
      license
      B
      quality
      D
      maintenance
      An MCP server that allows AI assistants to utilize human capabilities by sending requests to humans and receiving their responses through a Streamlit UI.
      7
      49
      MIT
    • F
      license
      B
      quality
      D
      maintenance
      MCP server that enables human-in-the-loop workflow in AI-assisted development tools by allowing users to provide direct feedback to AI agents without consuming additional premium requests.
      1
      1
      -