mcpField
I
InfoSys Cambridge Hackathon: Multi-Agent Fact-Checking for Social Media

InfoSys Cambridge Hackathon: Multi-Agent Fact-Checking for Social Media

github.com/denizakansoyFrameworks & SDKs4Updated 76w ago
Repository

About InfoSys Cambridge Hackathon: Multi-Agent Fact-Checking for Social Media

FactCheckBot is a multi-agent system built on MCP that identifies and verifies factual claims on social media. Each agent handles a specific reasoning task, from claim extraction to evidence retrieval and verdict generation, producing transparent and explainable conclusions.

Key features

  1. 01Extract and isolate factual claims from tweets
  2. 02Search verified sources like Wikipedia and arXiv
  3. 03Compare evidence and generate credibility scores
  4. 04Use crowdsourced consensus for breaking news
  5. 05Generate human-readable verdicts with reasoning
  6. 06Post corrections and clarifications as replies

Use cases

  1. 01Real-time detection of misinformation on social media
  2. 02Quick fact verification for breaking events
  3. 03Credibility assessment when traditional sources are unavailable
  4. 04Production of transparent, explainable fact-check conclusions

MCP support

Built on the MCP framework, FactCheckBot coordinates all agents through a shared memory structure, enabling dynamic collaboration and context-aware reasoning.

Details

From the project README

InfoSys Cambridge Hackathon: Multi-Agent Fact-Checking for Social Media

FactCheckBot is a multi-agent system designed to identify, verify, and respond to factual claims on social media platforms like Twitter (X). Built using the Model Context Protocol (MCP), the system decomposes the complex task of fact-checking into modular, intelligent agents — each responsible for a specific decision or reasoning step.

What It Does

Given a tweet or thread, FactCheckBot:

  1. Extracts factual claims
  2. Searches reliable external sources
  3. Evaluates the claim based on supporting or contradicting evidence
  4. Falls back on crowdsourced consensus for recent events
  5. Generates a concise, explainable verdict
  6. Posts a correction or clarification as a reply

This system is designed to combat misinformation with speed, context, and transparency.

Agent Roles

Agent NameDescription
ClaimExtractorAgentDetects and isolates factual claims from tweets
RetrieverAgentSearches verified sources (Wikipedia, arXiv, news APIs, etc.)
VerifierAgentCompares the claim to retrieved evidence and makes an initial judgment
CrowdSignalAgentIn breaking news cases, checks for regional consensus across social media
ExplainerAgentSummarizes the verdict in human-readable language, with reasoning
PosterAgentPosts a reply to the original tweet, including sources and explanations

All agents operate on a shared memory structure defined by the Model Context Protocol, allowing dynamic coordination and context-aware reasoning.

Example Workflow

Input Tweet:

“Massive explosion reported in Ankara right now.”

Agent Flow:
  1. ClaimExtractorAgent detects a location-based factual claim.
  2. RetrieverAgent finds no results in verified sources — too recent.
  3. CrowdSignalAgent clusters 10+ independent geolocated users in Ankara posting similar claims.
  4. VerifierAgent assigns a tentative credibility score with fallback reasoning.
  5. ExplainerAgent formats the reply:

    “🚨 Developing: This event is not yet reported by major outlets, but multiple users in Ankara are sharing photos and consistent details. Treat as likely true, pending confirmation. [Context]”

  6. PosterAgent publishes the reply.

Real-Time Consensus via CrowdSignalAgent

In cases where a claim cannot be confirmed through traditional sources (e.g., breaking news or remote events), the system uses CrowdSignalAgent to:

  • Search for tweets matching the claim
  • Filter for geolocated, domain-relevant, or independent sources
  • Evaluate consensus within a time window
  • Provide a fallback credibility score if traditional verification is unavailable

This helps prevent false negatives and enhances trustworthiness in time-sensitive cases.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.