A powerful skill that leverages Mixture of Agents (MoA) to synthesize high-quality responses from multiple LLM providers in parallel.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install multi-llm
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install multi-llm using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Multi-LLM Ensemble?
The Multi-LLM skill is designed for developers who need higher reliability and reasoning depth than a single model can provide. By utilizing the Mixture of Agents (MoA) pattern, it coordinates multiple proposer models like Claude, GPT-4o, and Gemini to generate initial responses, which are then refined by a synthesizer model. This approach ensures that Openclaw Skills can deliver superior results for complex tasks by aggregating the unique strengths of various top-tier language models.
The system is built for efficiency, querying all providers in parallel to minimize latency. It intelligently auto-detects available providers based on your environment variables, making it a seamless addition to any AI-driven workflow that requires the highest possible output quality.
Multi-LLM Ensemble Use Cases
- Complex reasoning and strategic planning tasks that require diverse perspectives.
- High-stakes decision making where accuracy and cross-verification are paramount.
- In-depth code reviews and bug analysis across different model architectures.
- Research summarization and synthesizing information from multiple competing sources.
- Improving trading signal quality or financial analysis accuracy where precision is critical.
How Multi-LLM Ensemble Works
- Parallel Querying: The skill triggers all available proposer models simultaneously using asynchronous calls to maximize speed.
- Model Execution: Individual models from providers like OpenAI, Anthropic, and Google generate independent answers to the prompt.
- Synthesis Phase: A designated synthesizer model (usually the most capable available) reviews all proposer outputs to combine the most accurate and relevant elements.
- Iterative Refinement: The process can optionally repeat for multiple rounds, further polishing the answer for maximum quality.
- Output Delivery: The final synthesized result is delivered via stdout or saved to a file in either plain text or structured JSON format.
Multi-LLM Ensemble Setup
To get started with this skill within the Openclaw Skills ecosystem, ensure you have the uv package manager installed and set up your API keys.
# Set your provider API keys
export ANTHROPIC_API_KEY="your_key"
export OPENAI_API_KEY="your_key"
export GEMINI_API_KEY="your_key"
# Run a multi-model query
uv run scripts/ensemble.py --prompt "Analyze the security implications of this code snippet"
Multi-LLM Ensemble Data Schema & Taxonomy
The Multi-LLM skill organizes its execution data based on the selected output format. It can produce simple text strings or detailed metadata objects.
| Data Type | Description |
|---|---|
text |
A clean string containing only the final synthesized response. |
json |
A structured object containing the final answer plus a list of responses from each individual proposer model. |
output_file |
An optional file (e.g., .txt or .json) where the synthesis results are persisted for later audit. |
Multi-LLM Ensemble Advanced Features
- Mixture of Agents (MoA) architecture for state-of-the-art response synthesis and error correction.
- Automatic provider detection which gracefully skips models with missing API keys without crashing the workflow.
- Configurable synthesis rounds (e.g.,
--rounds 2) to prioritize extreme response quality over speed. - Manual model override allowing users to specify exact proposer and synthesizer models for specialized tasks.
- Seamless local integration with Ollama for hosting private or open-source models alongside cloud providers.
SKILL.md
Loading
Related Openclaw Skills
Feishu Robot Creator
A specialized skill for creating Feishu bots and binding them to independent OpenClaw agents for role-based automation.
Grant Writing Coach
A specialized AI assistant designed to coach non-profits, researchers, and artists through every stage of the grant writing and funding acquisition process.
Shopify Store Optimizer
A senior-level AI audit tool that diagnoses Shopify store performance, conversion leaks, and technical debt to produce prioritized growth roadmaps.
Self-Healing AI Agent (skylv-self-healing-agent)
A sophisticated self-repair engine for AI agents that automatically detects, diagnoses, and fixes workflow failures using pattern recognition.
skylv-cost-guard
A powerful utility for monitoring, comparing, and optimizing AI API expenditure across multiple model providers.
Web3 Dev: Blockscout PRO API Integration
A specialized skill for building web3 applications that query indexed blockchain data and contract state across 100+ EVM chains via the Blockscout PRO API.
Featured*
