bias-detection-mcp 1.1.4

AI bias detection and fairness assessment for AI agents. Instant bias risk scan, demographic analysis, fairness metrics, mitigation strategies, EU AI Act Article 10 compliance. Zero-config free tier. Built by MEOK AI Labs.

Bias Detection MCP

Buy Starter — £29/mo

Signed attestations + unlimited audits + email support. 👉 Subscribe at meok.ai — instant HMAC signing key + Stripe-managed billing.

Free tier remains MIT-licensed and zero-config. Upgrade only when you need signed compliance artefacts for audit.

MCP server for bias detection mcp operations

Quick Install

Client	Install
Claude Desktop
Cursor
VS Code
Windsurf
Docker	docker run -p 8000:8000 bias-detection-mcp
pip	pip install bias-detection-mcp

Overview

Bias Detection MCP provides AI-powered tools via the Model Context Protocol (MCP).

Tools

Tool	Description
quick_scan	Describe an AI system in one sentence -> instant bias risk assessment. No API ke
detect_bias	Analyze text for demographic bias patterns, stereotyping, and unfair language.
fairness_metrics	Calculate fairness metrics from prediction data. Input format: comma-separated v
mitigation_recommendations	Get detailed remediation steps for a specific type of AI bias.
regulatory_check	Check bias requirements against EU AI Act Article 10 and NIST AI RMF MAP require

Installation

pip install meok-bias-detection-mcp

Usage with Claude Desktop

Add to your Claude Desktop MCP config (claude_desktop_config.json):

{
  "mcpServers": {
    "bias-detection-mcp": {
      "command": "python",
      "args": ["-m", "meok_bias_detection_mcp.server"]
    }
  }
}

Usage with FastMCP

from mcp.server.fastmcp import FastMCP

# This server exposes 5 tool(s) via MCP
# See server.py for full implementation

If this tool helps your compliance workflow, please star this repo — it helps other teams find it.

Wire it up — full stack

Pair this with the MEOK chain that turns one agent action into ONE signed compliance event:

1. bft-progress-council-mcp — anti-loop guardrail
2. agent-token-budget-mcp — hard spend cap
3. agent-prompt-injection-firewall-mcp — OWASP LLM01 scan
4. agent-audit-logger-mcp — hash-chained evidence
5. a2a-governance-bridge-mcp — fold N attestations → 1 signed event
6. agent-incident-relay-mcp — broadcast incidents to 5 regimes simultaneously

See meok.ai/mcp-stack for the full architecture and meok.ai/mcp-stack/demo for the live in-browser demo.

License

MIT © MEOK AI Labs

<<<<<<< Updated upstream

<script type="application/ld+json"> { "@context": "https://schema.org", "@type": "FAQPage", "mainEntity": [ { "@type": "Question", "name": "Is this MCP server free to use?", "acceptedAnswer": { "@type": "Answer", "text": "Yes. The free tier gives you 10 calls per day with no API key required. Pro tier is £79/mo for unlimited calls plus cryptographically signed attestations your auditor can verify independently." } }, { "@type": "Question", "name": "How does the signed attestation work?", "acceptedAnswer": { "@type": "Answer", "text": "Every Pro tier audit produces a HMAC-SHA256 signed certificate with a unique ID and a public verify URL. Your auditor pastes the cert into https://meok-attestation-api.vercel.app/verify and gets an independent valid/invalid response. No contact with MEOK required." } }, { "@type": "Question", "name": "Which MCP clients does this work with?", "acceptedAnswer": { "@type": "Answer", "text": "All standard MCP clients: Claude Desktop, Claude Code, Cursor, VS Code with MCP extension, Windsurf, Cline, and any custom MCP-compatible agent. Install via npx meok-setup or pip install for the underlying Python package." } }, { "@type": "Question", "name": "Can I install all MEOK governance MCPs at once?", "acceptedAnswer": { "@type": "Answer", "text": "Yes. Run npx meok-setup --pack governance to install all 10 governance MCPs and write the configs for Claude Desktop, Cursor, or Windsurf in one command." } }, { "@type": "Question", "name": "Is the regulation text authoritative?", "acceptedAnswer": { "@type": "Answer", "text": "Yes. MEOK syncs daily from the EUR-Lex Cellar SPARQL endpoint, the canonical EU regulation publication system. The text is verbatim with no LLM summarization. Every quote is auditor-defensible and includes the exact article number plus relevance score." } } ] } </script>

Stashed changes

Sister MCPs

Part of the MEOK Governance pack — designed to work together as a fleet. Install the whole pack with npx meok-setup --pack governance, or pick the ones you need:

• EU AI Act → uvx eu-ai-act-compliance-mcp · PyPI · GitHub
• DORA → uvx dora-compliance-mcp · PyPI · GitHub
• NIS2 → uvx nis2-compliance-mcp · PyPI · GitHub
• Cyber Resilience Act → uvx cra-compliance-mcp · PyPI · GitHub
• AI Bill of Materials → uvx ai-bom-mcp · PyPI · GitHub
• AI Incident Reporting → uvx ai-incident-reporting-mcp · PyPI · GitHub

Full catalogue + Anthropic Registry verify links: meok.ai/anthropic-registry

Protocol coverage + Universal PAYG

This MCP is part of MEOK's 47-MCP fleet that bridges every active agent-interop protocol and 30+ regulatory frameworks. See the full coverage matrix at meok.ai/protocols.

Agent interop protocols supported (8 live):

• ✅ MCP (Anthropic) — native
• ✅ A2A (Google + Linux Foundation, absorbed IBM ACP Sept 2025)
• ✅ IBM ACP — covered via A2A merge
• ◐ Stripe ACP (Agentic Commerce Protocol) — Q3 bridge via agent-commerce-protocol-mcp
• ◐ AP2 (Google Agent Payments) — partial via agent-commerce-payments-mcp
• ◐ x402 (Coinbase HTTP 402) — partial via api.meok.ai gateway
• → OASF / AGNTCY (Cisco Outshift + Linux Foundation) — Q3 bridge
• 👁 ANP (Cisco Agent Network) — watch-list

Pricing options:

Option	Price	Best for
Self-host (this MCP)	£0 — MIT	Devs
This MCP Starter	£29/mo	One-MCP teams
This MCP Pro	£79/mo	Production + 24h SLA
Universal PAYG	£29/mo + £0.0002/call	Spiky usage across many MCPs
Substrate bundle (this category)	£99-£499/mo	A whole pack
MEOK Universe	£1,499/mo	All 47 MCPs, 500K calls

Each tier above the free self-host adds HMAC-signed attestations verifiable at verify.meok.ai. Linux Foundation governance on the A2A spine means EU regulated buyers can deploy without vendor-lock-in objections.