MCP For Claude Marketplace
Agent Hallucination Cross-Checker

Agent Hallucination Cross-Checker MCP Connector for Claude

A+

A deterministic engine to audit agent outputs by measuring consensus, source validity, and semantic contradictions.

3 tools Official Updated Oct 1, 2026 Official Vinkius Partner
More Details Connect to Claude

This MCP server provides a deterministic verification engine designed to audit multiple agent outputs. It identifies hallucinations by measuring consensus, source validity, and semantic contradictions. Using tools like verify_claim_consensus, detect_hallucinations, and analyze_agreement_depth, it calculates fact consistency scores, detects contradictions between high-confidence claims, and determines the probability of hallucinations. It is built to distinguish between total agreement, partial agreement, and outright contradictions to ensure high-fidelity agent interactions.

hallucinationconsensusverificationauditreliability

3 tools expose this connector's capabilities to your AI agent.

analyze_agreement_depth

Breaks down the nature of agent interactions to distinguish between total agreement, partial agreement, and contradictions

detect_hallucinations

Identifies specific claims that are likely to be hallucinations based on probabilistic modeling and contradiction detection

verify_claim_consensus

Calculates the overall reliability of the provided agent outputs through consistency and source metrics

See how to talk to your AI agent using Agent Hallucination Cross-Checker.

Analyze these agent outputs for consistency: [{'agentId': 'A', 'claim': 'The sky is blue', 'confidence': 0.9, 'source': 'ref_1'}, {'agentId': 'B', 'claim': 'The sky is blue', 'confidence': 0.8, 'source': 'ref_1'}]

The fact consistency score is 1.0, as both agents provided identical claims with valid sources.

Check for contradictions in these claims: [{'agentId': 'A', 'claim': 'The temperature is rising', 'confidence': 0.9, 'source': 'ref_1'}, {'agentId': 'B', 'claim': 'The temperature is falling', 'confidence': 0.85, 'source': 'ref_1'}]

A contradiction was detected because the claims are semantic opposites and both agents reported confidence above 0.7.

Calculate the consensus for these outputs with a minimum quorum of 2.

The consensus report shows that the required quorum has been met for the verified facts.

The engine calculates a hallucination probability for each claim by analyzing source validity, cross-agent agreement, and confidence calibration.

Related Connectors

Prompt Distillation Calculator

Prompt Distillation Calculator MCP

3 tools Official

Calculate efficiency, quality, and cost of prompt distillation.

A+ View details →
Agent Config Drift Detector

Agent Config Drift Detector MCP

3 tools Official

Detects unauthorized changes to agent configurations by comparing SHA-256 hashes.

A+ View details →
Conversation Turn Fairness Enforcer

Conversation Turn Fairness Enforcer MCP

3 tools Official

Prevents agent domination in multi-agent conversations by monitoring turn frequency.

A+ View details →
Helicone (LLM Observability)

Helicone (LLM Observability) MCP

10 tools Official

Monitor LLM usage via Helicone — track requests, analyze costs, measure latency, and manage prompts.

A+ View details →