Integrates with Google AI (Gemini) models to provide multi-agent thinking augmentation through parallel critique analysis using Gemini 2.5 Flash for critiques and Gemini 2.5 Pro for synthesis
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Elrond MCP analyze our plan to launch a new AI-powered customer feedback system"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Model Context Protocol (MCP) server that provides hierarchical LLM critique and synthesis for enhanced decision-making and idea evaluation.
Preview Software: This is experimental software in active development and is not intended for production use. Features may change, break, or be removed without notice. Use at your own risk.
Elrond MCP implements a multi-agent thinking augmentation system that analyzes proposals through three specialized critique perspectives (positive, neutral, negative) and synthesizes them into comprehensive, actionable insights. This approach helps overcome single-model biases and provides more thorough analysis of complex ideas.
Related MCP server: VerifiMind PEAS
Parallel Critique Analysis: Three specialized agents analyze proposals simultaneously from different perspectives
Structured Responses: Uses Pydantic models and instructor library for reliable, structured outputs
Google AI Integration: Leverages Gemini 2.5 Flash for critiques and Gemini 2.5 Pro for synthesis
MCP Compliance: Full Model Context Protocol support for seamless integration with AI assistants
Comprehensive Analysis: Covers feasibility, risks, benefits, implementation, stakeholder impact, and resource requirements
Consensus Building: Identifies areas of agreement and disagreement across perspectives
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ Positive │ │ Neutral │ │ Negative │
│ Critique │ │ Critique │ │ Critique │
│ Agent │ │ Agent │ │ Agent │
│ │ │ │ │ │
│ Gemini 2.5 │ │ Gemini 2.5 │ │ Gemini 2.5 │
│ Flash │ │ Flash │ │ Flash │
└─────────┬───────┘ └─────────┬───────┘ └─────────┬───────┘
│ │ │
│ │ │
└──────────────────────┼──────────────────────┘
│
▼
┌─────────────────────────┐
│ Synthesis Agent │
│ │
│ Gemini 2.5 Pro │
│ │
│ │
│ Consensus + Summary │
└─────────────────────────┘Python 3.13 or higher
Google AI API key (get one at Google AI Studio)
Clone the repository:
git clone <repository-url>
cd elrond-mcpInstall dependencies:
# Using uv (recommended)
uv sync --dev --all-extras
# Or using pip
pip install -e .[dev]Configure API key:
export GEMINI_API_KEY="your-gemini-api-key-here"
# Or create a .env file
echo "GEMINI_API_KEY=your-gemini-api-key-here" > .env# Using uv
uv run python main.py
# Using MCP CLI (if installed)
mcp dev elrond_mcp/server.py# Direct execution
python main.py
# Or via package entry point
elrond-mcpInstall for Claude Desktop:
mcp install elrond_mcp/server.py --name "Elrond Thinking Augmentation"Manual Configuration: Add to your Claude Desktop MCP settings:
{
"elrond-mcp": {
"command": "python",
"args": ["/path/to/elrond-mcp/main.py"],
"env": {
"GEMINI_API_KEY": "your-api-key-here"
}
}
}Analyze any proposal through multi-perspective critique:
Use the "consult_the_council" tool with this proposal:
# Project Alpha: AI-Powered Customer Service
## Overview
Implement an AI chatbot to handle 80% of customer service inquiries, reducing response time from 2 hours to 30 seconds.
## Goals
- Reduce operational costs by 40%
- Improve customer satisfaction scores
- Free up human agents for complex issues
## Implementation
- Deploy GPT-4 based chatbot
- Integrate with existing CRM
- 3-month rollout plan
- $200K initial investmentMonitor the health and configuration of the thinking augmentation system:
Use the "check_system_status" tool to verify:
- API key configuration
- Model availability
- System healthEach critique agent provides:
Executive Summary: Brief overview of the perspective
Structured Analysis:
Feasibility assessment
Risk identification
Benefit analysis
Implementation considerations
Stakeholder impact
Resource requirements
Key Insights: 3-5 critical observations
Confidence Level: Numerical confidence (0.0-1.0)
The synthesis agent provides:
Executive Summary: High-level recommendation
Consensus View:
Areas of agreement
Areas of disagreement
Balanced assessment
Critical considerations
Recommendation: Overall guidance
Next Steps: Concrete action items
Uncertainty Flags: Areas needing more information
Overall Confidence: Synthesis confidence level
elrond-mcp/
├── elrond_mcp/
│ ├── __init__.py
│ ├── server.py # MCP server implementation
│ ├── agents.py # Critique and synthesis agents
│ ├── client.py # Centralized Google AI client management
│ └── models.py # Pydantic data models
├── scripts/ # Development scripts
│ └── check.sh # Quality check script
├── tests/ # Test suite
├── main.py # Entry point
├── pyproject.toml # Project configuration
└── README.md# Using uv
uv run pytest
# Using pip
pytest# Format and lint code
uv run ruff format .
uv run ruff check --fix .
# Type checking
uv run mypy elrond_mcp/For convenience, use the provided script to run all quality checks:
# Run all quality checks (lint, format, test)
./scripts/check.shThis script will:
Sync dependencies
Run Ruff linter with auto-fix
Format code with Ruff
Execute the full test suite
Perform final lint check
Provide a pre-commit checklist
GEMINI_API_KEY: Required Google AI API key
LOG_LEVEL: Logging level (default: INFO)
Critique Agents: gemini-2.5-flash
Synthesis Agent: gemini-2.5-pro
Models can be customized by modifying the agent initialization in agents.py.
API Key Not Found
Error: Google AI API key is requiredSolution: Set the GEMINI_API_KEY environment variable
Empty Proposal Error
Error: Proposal cannot be emptySolution: Ensure your proposal is at least 10 characters long
Model Rate Limits
Error: Rate limit exceededSolution: Wait a moment and retry, or check your Google AI quota
Validation Errors
ValidationError: ...Solution: The LLM response didn't match expected structure. This is usually temporary - retry the request
Enable debug logging:
export LOG_LEVEL=DEBUG
export GEMINI_API_KEY=your-api-key-here
python main.pyCheck system status:
# Use the check_system_status tool to verify configurationFork the repository
Create a feature branch
Make your changes
Add tests for new functionality
Run the test suite
Submit a pull request
See LICENSE
For issues and questions:
Check the troubleshooting section above
Review the logs for detailed error information
Open an issue on the repository
Support for additional LLM providers (OpenAI, Anthropic)
Custom critique perspectives and personas
Performance optimization and caching
Advanced synthesis algorithms
Check the status of the thinking augmentation system.
Returns: Dictionary containing system status information including API key availability, model configurations, and health status
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It discloses that the tool returns a dictionary with specific status information (API key availability, model configurations, health status), which adds useful context beyond a basic read operation. However, it doesn't cover other behavioral aspects like error handling, performance, or prerequisites, leaving gaps.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is appropriately sized and front-loaded, with the core purpose stated first and return details following. It uses two sentences efficiently, though the second sentence could be slightly more concise (e.g., by integrating the return details into the first sentence).
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity (0 parameters, no output schema, no annotations), the description is adequate but has clear gaps. It explains the return content well, but without annotations or output schema, it lacks details on format, error cases, or integration with the sibling tool, making it minimally viable.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has 0 parameters, and schema description coverage is 100%, so no parameter information is needed. The description appropriately doesn't discuss parameters, earning a baseline score of 4 for not adding unnecessary details.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose as checking the status of a specific system ('thinking augmentation system'), which is a specific verb+resource combination. However, it doesn't explicitly differentiate from the sibling tool 'consult_the_council', which might be related but isn't described here, preventing a perfect score.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides no guidance on when to use this tool versus alternatives. It doesn't mention the sibling tool 'consult_the_council' or any other context for usage decisions, leaving the agent with no explicit or implied usage instructions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Analyze a proposal using hierarchical LLM critique and synthesis.
This tool runs three specialized critique agents (positive, neutral, negative) in parallel to analyze the proposal, then synthesizes their perspectives into a comprehensive analysis with recommendations.
Args: proposal: Markdown-formatted proposal outlining the salient points of the solution to be analyzed
Returns: Complete analysis including all critiques and synthesis with consensus view, recommendations, and next steps
| Name | Required | Description | Default |
|---|---|---|---|
| proposal | Yes |
| Name | Required | Description |
|---|---|---|
| critiques | Yes | All critique responses (positive, neutral, negative) |
| synthesis | Yes | Final synthesis of all perspectives |
| original_proposal | Yes | The original proposal that was analyzed |
| processing_metadata | No | Metadata about the processing (timestamps, model versions, etc.) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It effectively describes the tool's behavior: running three specialized critique agents in parallel, analyzing the proposal from positive/neutral/negative perspectives, and synthesizing results. It mentions the parallel execution pattern and multi-perspective approach, which are valuable behavioral insights beyond basic function. However, it doesn't address potential limitations like processing time, token limits, or error conditions.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is well-structured and front-loaded with the core purpose in the first sentence. Each subsequent sentence adds specific value: explaining the three-agent architecture, describing the synthesis process, documenting the parameter, and outlining return values. There's no wasted text, and the information is presented in a logical flow from high-level purpose to implementation details.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (multi-agent analysis with synthesis), no annotations, and the presence of an output schema, the description provides good coverage. It explains the analysis methodology, parameter requirements, and return content. The output schema existence means the description doesn't need to detail return structure. However, for a complex tool with no annotations, it could benefit from mentioning potential constraints or limitations.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema has 0% description coverage, so the description must compensate. It provides meaningful context for the single parameter by specifying it should be 'Markdown-formatted' and should outline 'salient points of the solution to be analyzed.' This adds valuable semantic information beyond the bare schema type. However, it doesn't provide examples or more detailed formatting requirements.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose with specific verbs ('analyze', 'runs', 'synthesizes') and resources ('proposal', 'three specialized critique agents'). It distinguishes itself from the only sibling tool (check_system_status) by focusing on proposal analysis rather than system monitoring. The description explains the hierarchical LLM critique and synthesis process, making the purpose unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage context by stating it analyzes 'a proposal' and outlines the specific process, but provides no explicit guidance on when to use this tool versus alternatives. Since there's only one sibling tool (check_system_status) with a completely different function, the lack of comparative guidance is less critical, but no explicit when/when-not instructions are provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
check_system_statusconsult_the_councilThe two tools have completely distinct purposes: one checks system status, while the other analyzes proposals through hierarchical critique. There is no overlap in functionality or potential for confusion between monitoring system health and conducting proposal analysis.
Both tools follow a consistent verb_noun naming pattern: check_system_status and consult_the_council. The naming is clear, descriptive, and follows the same convention throughout the toolset.
With only two tools, this server feels severely under-equipped for what appears to be a thinking augmentation system. The tools cover basic status checking and proposal analysis, but there are likely many missing operations for a comprehensive augmentation system (e.g., configuration management, history tracking, different analysis modes).
The tool surface is significantly incomplete for a thinking augmentation system. While the two tools provide some functionality, there are obvious gaps: no way to configure the system, no historical analysis tracking, no ability to modify critique parameters, and no integration with external data sources. The tools feel like isolated endpoints rather than a complete system.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Multi-LLM council: 25+ frontier models in parallel, consensus scoring, verdict-first code review.
Multi-agent governance: task orchestration, compliance, decision validation, and ML predictions.
Multi-expert decision intelligence with transparent synthesis and auditable workflows.
Multi-LLM AI Research & Analysis — smart routing, consensus analysis, due diligence reports
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/dogonthehorizon/elrond-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
