mmcp
by RagavRida
DAG-based multi-model orchestration with RL routing and multi-verifier voting
Overview
Orchestrates multiple language models and agents as a directed acyclic graph (DAG) to solve complex tasks. Uses RL-based routing, multi-verifier voting, and an agent mesh so results can be validated and the pipeline can self-improve over time. Provides TypeScript and Python packages for plugging in OpenAI, Anthropic, Gemini, DeepSeek and custom model adapters, enabling Semantic Capability Matching Pattern to align models and tasks.
Key Benefits
Target Use Cases
Engineering teams building multi-model pipelines who need structured routing, verification, and self-improvement across heterogeneous LLMs. Consider applying the Market-Based Coordination Pattern to optimize resource use and coordination, alongside Evaluation-Driven Development (EDDOps).
Use Cases
- Route user requests across specialist models and aggregate verified outputs
- Run ensemble verification pipelines with voting and multi-verifier checks
- Continuously improve routing policies with reinforcement learning feedback
- Prototype multi-model agent meshes that combine commercial and custom models