Skip to main content
Back to challenges
Challenge

Build MCP-Enabled Social Media Policy Enforcement Agents

Companies are increasing social media monitoring due to reputational risks from employee posts. This challenge involves building an advanced multi-agent system that autonomously monitors public social media feeds for potential employee policy violations. The system will leverage a graph-based workflow to analyze content, interpret company policies (retrieved via RAG), and flag potential issues, while adapting its reasoning budget based on the sensitivity of the content or the severity of the potential violation. You will design and implement a sophisticated LangGraph-based agent network where a 'Monitor Agent' feeds data to a 'Policy Interpretation Agent' and a 'Risk Assessment Agent'. These agents will communicate using a defined protocol, integrating with external enterprise policy databases via MCP for real-time policy lookup and dynamically adjusting their processing depth (thinking budget) to balance efficiency and accuracy. The system must provide actionable insights and prioritize alerts for human review, ensuring compliance while minimizing false positives.

Special Purpose AgentsHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Design and deploy a LangGraph-based multi-agent system with Claude Opus 4.5 for adaptive social media policy enforcement, utilizing MCP for tool integration and extended thinking.

Delivery guide

How work is evaluated

Evaluation

The evaluation will assess the system's ability to accurately identify policy violations, its adaptability in reasoning, and the efficiency of its agent communication within the LangGraph framework. Emphasis will be placed on the correct implementation of MCP for tool calls and dynamic budget allocation.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Learning goals

What you should walk away with

  • Design a graph-based multi-agent workflow for social media monitoring using LangGraph.

  • Develop agents capable of extended thinking with adaptive reasoning budgets.

  • Implement MCP for secure and efficient tool integration with external policy databases.

  • Orchestrate agent-to-agent communication for collaborative policy enforcement.

  • Utilize RAG with LlamaIndex for dynamic policy lookup and contextual understanding.

Resources and assets

Reference links and supporting material

Dataset notes

Sample data for 1 tasks

Challenge setup

The evaluation will assess the system's ability to accurately identify policy violations, its adaptability in reasoning, and the efficiency of its agent communication within the LangGraph framework. Emphasis will be p...

Challenge input

{ 'social_media_posts': [{'id': 'str', 'text': 'str', 'author_id': 'str'}], 'company_policies': [{'id': 'str', 'text': 'str'}] }

Agent execution

No agent tool is listed for this challenge.

Evaluated output

{ 'alerts': [{'post_id': 'str', 'violation_type': 'str', 'reasoning': 'str', 'severity': 'low|medium|high', 'adaptive_budget_used': 'float'}] }

Evaluation checks
  • Successfully identifies all planted policy violations in the sample input.
  • Demonstrates successful MCP calls for policy lookup in the agent's internal logs.
  • Shows evidence of dynamic reasoning budget adjustment based on detected severity.
Configured evaluation evidence
  • Precision target: 0.95
Configured runtime
  • Python execution harness

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

No tool records are attached.

Evaluation contract

  • Successfully identifies all planted policy violations in the sample input.
  • Demonstrates successful MCP calls for policy lookup in the agent's internal logs.
  • Shows evidence of dynamic reasoning budget adjustment based on detected severity.

Metrics

  • Precision · Target 0.95
  • Recall · Target 0.95
  • AverageAdaptiveBudget · Target 0.5

Workflow tasks

  1. Task 1

    Input: { 'social_media_posts': [{'id': 'str', 'text': 'str', 'author_id': 'str'}], 'company_policies': [{'id': 'str', 'text': 'str'}] }

    Output: { 'alerts': [{'post_id': 'str', 'violation_type': 'str', 'reasoning': 'str', 'severity': 'low|medium|high', 'adaptive_budget_used': 'float'}] }

    Execution harness: Python

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build MCP-Enabled Social Media Policy Enforcement Agents