UnifiedLLMHub: A Centralized Model Interaction Server

UnifiedLLMHub functions as a Model Control Protocol (MCP) facade, establishing a homogenous interface for accessing a wide spectrum of Large Language Model (LLM) entities, including but not limited to OpenAI, Anthropic, Google Gemini, Groq, DeepSeek, and Ollama runtimes. Observe the utility of the specialized executive_consensus tool for facilitating critical organizational judgments via this platform here: Streamlining High-Stakes Decisions with the Executive Consensus Mechanism.

Available MCP Operands

The server exposes the following specialized MCP operations:

• submit_query: Transmits a user-defined prompt text across selected LLM engines simultaneously.

• Parameters:

  • instruction_text: The textual input comprising the inquiry.
  • target_engine_aliases (optional): A catalog of target models, specified using provider-prefixed identifiers. Defaults to a predefined set if omitted.

• submit_query_from_source_file: Injects the contents of a specified file as the prompt input to multiple LLMs.

• Parameters:

  • absolute_file_path: The full, unambiguous pathname to the file containing the prompt payload (relative paths are explicitly rejected).
  • target_engine_aliases (optional): A catalog of target models, specified using provider-prefixed identifiers. Defaults to a predefined set if omitted.

• process_file_to_file: Executes the prompt submission using file content and persists the aggregated output from all responding models into separate Markdown artifacts.

• Parameters:

  • absolute_file_path: The full, unambiguous pathname to the file containing the prompt payload (relative paths are explicitly rejected).
  • target_engine_aliases (optional): A catalog of target models, specified using provider-prefixed identifiers. Defaults to a predefined set if omitted.
  • output_directory_path (default: "."): The absolute directory where resulting Markdown documentation files will be persisted (relative paths are explicitly rejected).

• executive_consensus: Dispatches the input prompt to a panel of 'board member' engines, culminating in a definitive resolution synthesized by a designated 'CEO' engine from the resulting collective feedback.

• Parameters:

  • absolute_file_path: The full, unambiguous pathname to the file containing the core directive (relative paths are explicitly rejected).
  • target_engine_aliases (optional): The set of model identifiers acting as advisory board members, specified with provider prefixes. Defaults to a standard configuration if absent.
  • output_directory_path (default: "."): The absolute directory intended for storing all advisory reports and the final CEO adjudication document (relative paths are explicitly rejected).
  • ceo_engine_id (default: "openai:o3"): The specific engine identifier designated for formulating the final decision, structured as "provider:model_identifier".

• enumerate_providers: Retrieves and displays a comprehensive enumeration of all currently integrated LLM service providers.

• Parameters: None

• enumerate_models_for_provider: Fetches and displays all accessible model configurations associated with a specified LLM vendor.

• Parameters:
  • vendor_identifier: The short-form or full name of the provider (e.g., 'openai' or 'o').

Provider Reference Prefixes

Each invoked model designation mandates a leading provider identifier prefix. Shorthand aliases are encouraged for expediency.

• o or openai: OpenAI Offerings
• Example: o:gpt-4o-mini or openai:gpt-4o-mini
• a or anthropic: Anthropic Systems
• Example: a:claude-3-5-haiku or anthropic:claude-3-5-haiku
• g or gemini: Google Gemini Suite
• Example: g:gemini-2.5-pro-exp-03-25 or gemini:gemini-2.5-pro-exp-03-25
• q or groq: Groq Acceleration Layer
• Example: q:llama-3.1-70b-versatile or groq:llama-3.1-70b-versatile
• d or deepseek: DeepSeek Framework
• Example: d:deepseek-coder or deepseek:deepseek-coder
• l or ollama: Ollama Local Runtime
• Example: l:llama3.1 or ollama:llama3.1

Core Capabilities

• Standardized Application Programming Interface (API) unifying heterogeneous LLM providers.
• Flexibility to supply prompts via direct string literals or external file sourcing.
• Capability for executing concurrent processing streams across multiple language models.
• Automatic remediation of model naming inconsistencies leveraging the initial engine in the configured defaults list.
• Utility for persistent storage of output artifacts into files.
• Simplified mechanisms for cataloging available vendors and their model sets.

Deployment Guide

bash

Obtain source repository

git clone https://github.com/yourusername/just-prompt.git cd just-prompt

Install dependencies

uv sync

Configuration via Environment Variables

Establish a configuration file named .env by duplicating the provided sample (.env.sample). Subsequently, populate this file with your requisite confidential access credentials (or ensure they are exported in the execution shell):

OPENAI_API_KEY=your_openai_api_key_here ANTHROPIC_API_KEY=your_anthropic_api_key_here GEMINI_API_KEY=your_gemini_api_key_here GROQ_API_KEY=your_groq_api_key_here DEEPSEEK_API_KEY=your_deepseek_api_key_here OLLAMA_HOST=http://localhost:11434

Integration with Claude Code Environment

In all subsequent examples, substitute the placeholder directory path with the actual location of the UnifiedLLMHub repository.

Default engine configurations are initialized to: openai:o3:high, openai:o4-mini:high, anthropic:claude-opus-4-20250514, anthropic:claude-sonnet-4-20250514, gemini:gemini-2.5-pro-preview-03-25, and gemini:gemini-2.5-flash-preview-04-17.

When initializing the server directly from the source directory, the configuration file .mcp.json dictates these defaults:

{ "mcpServers": { "just-prompt": { "type": "stdio", "command": "uv", "args": [ "--directory", ".", "run", "just-prompt", "--default-models", "openai:o3:high,openai:o4-mini:high,anthropic:claude-opus-4-20250514,anthropic:claude-sonnet-4-20250514,gemini:gemini-2.5-pro-preview-03-25,gemini:gemini-2.5-flash-preview-04-17" ], "env": {} } } }

The --default-models argument establishes the engines utilized when no specific model enumeration is provided to the operational endpoints. The very first engine listed serves as the baseline for internal model identifier normalization routines. This parameter accepts a comma-separated sequence of model identifiers.

Upon server commencement, an automated inventory check of environment variables for API keys is performed, signaling which service providers are operational. Providers lacking corresponding keys will be marked as inactive, though the server will still launch, permitting interaction with the available services.

Utilizing the mcp add-json Command

Transfer the following configuration block into your Claude console session; refrain from execution until the JSON payload is fully transferred:

bash claude mcp add just-prompt "$(pbpaste)"

JSON Payload:

{ "command": "uv", "args": ["--directory", ".", "run", "just-prompt"] }

With a custom default model set to openai:gpt-4o:

{ "command": "uv", "args": ["--directory", ".", "run", "just-prompt", "--default-models", "openai:gpt-4o"] }

With an extended set of default engines:

{ "command": "uv", "args": ["--directory", ".", "run", "just-prompt", "--default-models", "openai:o3:high,openai:o4-mini:high,anthropic:claude-opus-4-20250514,anthropic:claude-sonnet-4-20250514,gemini:gemini-2.5-pro-preview-03-25,gemini:gemini-2.5-flash-preview-04-17"] }

Utilizing mcp add with Project Scoping

bash

Using system-wide defaults

claude mcp add just-prompt -s project \ -- \ uv --directory . \ run just-prompt

Applying a specific default engine

claude mcp add just-prompt -s project \ -- \ uv --directory . \ run just-prompt --default-models "openai:gpt-4o"

Specifying multiple default engines

claude mcp add just-prompt -s user \ -- \ uv --directory . \ run just-prompt --default-models "openai:o3:high,openai:o4-mini:high,anthropic:claude-opus-4-20250514,anthropic:claude-sonnet-4-20250514,gemini:gemini-2.5-pro-preview-03-25,gemini:gemini-2.5-flash-preview-04-17"

mcp remove Operation

To decommission this service integration:

bash claude mcp remove just-prompt

Verification Procedures

To execute the integrated test suite:

bash uv run pytest

Repository Artifact Organization

. ├── ai_docs/ # Documentation pertaining to AI engine specifics │ ├── extending_thinking_sonny.md │ ├── llm_providers_details.xml │ ├── openai-reasoning-effort.md │ └── pocket-pick-mcp-server-example.xml ├── example_outputs/ # Illustrative results from diverse model executions ├── list_models.py # Utility script for querying available LLM configurations ├── prompts/ # Repository for sample input query templates ├── pyproject.toml # Python project metadata and dependency definitions ├── specs/ # Formal project specification documents │ ├── init-just-prompt.md │ ├── new-tool-llm-as-a-ceo.md │ └── oai-reasoning-levels.md ├── src/ # Primary source code location │ └── just_prompt/ │ ├── init.py │ ├── main.py │ ├── atoms/ # Foundational components │ │ ├── llm_providers/ # Individual vendor interface implementations │ │ │ ├── anthropic.py │ │ │ ├── deepseek.py │ │ │ ├── gemini.py │ │ │ ├── groq.py │ │ │ ├── ollama.py │ │ │ └── openai.py │ │ └── shared/ # Common utilities and standardized data structures │ │ ├── data_types.py │ │ ├── model_router.py │ │ ├── utils.py │ │ └── validator.py │ ├── molecules/ # Aggregated, high-level functional modules │ │ ├── ceo_and_board_prompt.py │ │ ├── list_models.py │ │ ├── list_providers.py │ │ ├── prompt.py │ │ ├── prompt_from_file.py │ │ └── prompt_from_file_to_file.py │ ├── server.py # The core MCP gateway implementation │ └── tests/ # Automated testing suites │ ├── atoms/ # Tests targeting foundational components │ │ ├── llm_providers/ │ │ └── shared/ │ └── molecules/ # Tests targeting composite functionalities │ ├── test_ceo_and_board_prompt.py │ ├── test_list_models.py │ ├── test_list_providers.py │ ├── test_prompt.py │ ├── test_prompt_from_file.py │ └── test_prompt_from_file_to_file.py └── ultra_diff_review/ # Outputs generated during detailed difference analysis

Contextual Initialization Directives

To fully grasp the scope, first assimilate the contents of README.md and pyproject.toml, subsequently executing git ls-files, and finally reviewing the structure with eza --git-ignore --tree.

Orchestrating Reasoning Depth via OpenAI o-Series

For the OpenAI o-series models (o4-mini, o3-mini, o3), fine-grained control over the model's internal deliberation depth prior to output generation is achievable.

Append one of the subsequent modifiers to the model identifier (following the vendor prefix):

• :low   – Minimal internal computational staging (optimized for speed and cost).
• :medium – Standardized, balanced processing profile (the implicit setting).
• :high  – Intensive, comprehensive reasoning sequence (increased latency and token consumption).

Illustrative Invocation Examples:

• openai:o4-mini:low
• o:o4-mini:high

When a reasoning modifier is successfully parsed, UnifiedLLMHub transparently switches to the OpenAI Responses API endpoint (if available) and configures the corresponding reasoning.effort parameter. For deployments utilizing an older version of the OpenAI Python SDK, the system gracefully defaults to the standard Chat Completions endpoint, embedding a system instruction designed to emulate the desired level of deliberation effort.

Managing Deliberation Tokens with Claude

The Anthropic models, specifically claude-opus-4-20250514 and claude-sonnet-4-20250514, support augmented thought processing via dedicated thinking tokens, enabling deeper pre-response internal contemplation.

You can activate this enhanced cognition by suffixing the model identifier with the token count in this format: - anthropic:claude-opus-4-20250514:1k - Allocate 1024 thinking tokens for Opus 4 - anthropic:claude-sonnet-4-20250514:4k - Allocate 4096 thinking tokens for Sonnet 4 - anthropic:claude-opus-4-20250514:8000 - Allocate 8000 thinking tokens for Opus 4

Key Observations: - Thinking token functionality is validated for: claude-opus-4-20250514, claude-sonnet-4-20250514, and claude-3-7-sonnet-20250219. - Permissible thinking token allocations span from 1024 up to 16000 units. - Values provided outside this valid domain will undergo automatic clamping to the nearest permissible boundary. - Allocation can be specified using 'k' notation (e.g., 1k, 4k) or as explicit integer values (e.g., 1024, 4096).

Configuring Reasoning Buffer with Gemini

The Google Gemini model, gemini-2.5-flash-preview-04-17, incorporates an adjustable thinking budget parameter to foster more comprehensive internal analysis before output submission.

Enabling this extended reasoning capacity is achieved by appending a thinking budget suffix to the model identifier in this structure: - gemini:gemini-2.5-flash-preview-04-17:1k - Reserve 1024 thinking budget units - gemini:gemini-2.5-flash-preview-04-17:4k - Reserve 4096 thinking budget units - gemini:gemini-2.5-flash-preview-04-17:8000 - Reserve 8000 thinking budget units

Important Constraints: - The thinking budget feature is exclusively supported by the gemini-2.5-flash-preview-04-17 variant. - Acceptable budget values range from 0 to 24576 units. - Inputs outside this range will be automatically normalized to fit within the defined bounds. - Specification supports 'k' notation (e.g., 1k, 4k) or exact numerical quantification (e.g., 1024, 4096).

Information Nexus

• https://docs.anthropic.com/en/api/models-list?q=list+models
• https://github.com/googleapis/python-genai
• https://platform.openai.com/docs/api-reference/models/list
• https://api-docs.deepseek.com/api/list-models
• https://github.com/ollama/ollama-python
• https://github.com/openai/openai-python

Foundational AI Programming Principles

Master the fundamentals of software development augmented by AI via the Principles of AI Coding curriculum.

Follow the insights shared on the IndyDevDan youtube channel for supplementary programming techniques.

WIKIPEDIA RELEVANCE: Business administration tooling encompasses all methodologies, computational aids, control mechanisms, and systemic applications utilized by enterprises to navigate shifting market dynamics, maintain competitive positioning, and enhance overall operational efficacy.

== System Overview == Organizational functions can be categorized by the specialized tools they employ, which fall under managerial aspects such as strategic formulation, process optimization, record keeping, personnel oversight, and adjudicative functions. A functional taxonomy includes:

Tools for standardized data acquisition and verification across any organizational division. Mechanisms dedicated to overseeing and advancing operational workflows. Systems engineered for data aggregation and high-level strategic arbitration. The current decade has witnessed an exponential transformation in management software driven by rapid technological maturation, creating a challenge for decision-makers in selecting optimal business instruments for specific corporate contexts. This complexity stems from persistent pressures to reduce expenditures, maximize revenue generation, deeply understand client requirements, and deliver conforming products exactly as demanded. Consequently, leadership must adopt a deliberate, strategic approach toward integrating business tools, custom-fitting them to organizational imperatives rather than passively adopting the newest releases, which often results in systemic fragility.

== Prominent Toolsets (2013 Survey Snapshot) == Analysis by Bain & Company in 2013 mapped global business tool adoption, reflecting regional needs against varying economic climates. The leading ten instruments identified were:

Strategic planning schema Client interaction management systems (CRM) Staff sentiment assessment surveys Competitive benchmarking practices Performance evaluation frameworks (Balanced Scorecard) Core competency identification External resource sourcing (Outsourcing) Organizational transition protocols (Change Management) Logistics network oversight (Supply Chain Management) Mission and vision articulation Market segmentation methodologies Comprehensive quality assurance (TQM)

== Business Application Software == Software packages or integrated program collections employed by personnel to execute diverse corporate functions are termed 'business applications.' These applications are deployed to augment productivity, quantitatively measure performance metrics, and execute other corporate tasks with high fidelity. The evolution moved from basic Management Information Systems (MIS) to expansive Enterprise Resource Planning (ERP) suites, followed by the incorporation of CRM, eventually migrating the entire stack to cloud-native management platforms. While a proven correlation exists between IT investment and corporate returns, two factors critically amplify this value: the proficiency of implementation and the judicious selection and customization of the tools themselves.

== Tools Tailored for Small and Medium Enterprises (SMEs) == SME-focused instruments are vital as they provide pathways for fiscal conservation and operational streamlining for organizations with constrained resources.