mmcp-core 2.2.0

MMCP — Multi-Model Collaboration Pipeline. Auto-routes tasks to Claude, GPT, Gemini, DeepSeek with RL routing, cost optimization, and MCP server support.

MMCP Python SDK

Self-orchestrating multi-model AI pipeline — auto-routes tasks to the best AI model with smart routing, streaming, and audit trails.

Install

pip install mmcp-core

# With LangChain/LangGraph support
pip install mmcp-core[langchain]

# With Cloud server (PostgreSQL + Stripe)
pip install mmcp-core[cloud]

# Everything
pip install mmcp-core[all]

Quick Start — CLI

# Interactive mode with smart routing
mmcp run

# Autonomous pipeline — MMCP plans and executes
mmcp auto "Research Python best practices and write a summary"

# Direct pipeline commands
mmcp chain "Explain DAGs" -r architect,reviewer --openrouter
mmcp parallel "Analyze market trends" -f analyst,creative,critic -M synthesizer
mmcp verify "Evaluate React vs Vue" -p expert -c challenger -s synthesizer
mmcp shard "Write a report on AI" -r analyst -n 3 -M editor

Quick Start — Python SDK

import asyncio
from mmcp_core import MMCPOrchestrator, RoleBasedRouter, MemoryStore

async def main():
    orc = MMCPOrchestrator({
        "router": RoleBasedRouter({
            "architect": {"model_id": "claude-haiku-4-5-20251001"},
            "reviewer":  {"model_id": "claude-haiku-4-5-20251001"},
        }),
        "store": MemoryStore(),
    })

    result = await orc.run_chain(
        "Explain the observer pattern in Python.",
        ["architect", "reviewer"]
    )

    print(f"✅ {result.output}")
    print(f"🪙 Tokens: {result.total_tokens}")
    print(f"💰 Cost: ${result.total_cost_usd:.6f}")

asyncio.run(main())

Streaming Execution (v1.1.0)

from mmcp_core.stream_executor import stream_execute
from mmcp_core.planner import plan_task

plan = await plan_task("Research AI trends", api_key="sk-or-...")

async for event in stream_execute(plan):
    if event["type"] == "step_token":
        print(event["content"], end="", flush=True)
    elif event["type"] == "step_done":
        print(f"\n✓ Step {event['step']} ({event['tokens']} tokens)")
    elif event["type"] == "plan_done":
        print(f"\nTotal: {event['total_tokens']} tokens, ${event['total_cost']}")

v1.1.0 — What's New

Feature	Description
🧠 Autonomous Mode	mmcp auto — AI plans and executes multi-step tasks
🔀 Smart Routing	Auto-selects model per task (Gemini, GPT-4o, Claude, DeepSeek, Llama)
📡 Streaming	Real-time SSE output during pipeline execution
🏗️ CLI Package	Modular cli/ package (7 modules, was 1,299-line monolith)
🐘 PostgreSQL	Production-ready DB layer (SQLite for dev, PostgreSQL for prod)
💳 Stripe	Subscription billing for Pro/Team plans
✅ Test Suite	84 automated tests with GitHub Actions CI/CD

DAG Operations

Operation	Signature	Description
fork	1 → N	Spawn parallel sub-contexts
merge	N → 1	Combine parent outputs
handoff	1 → 1	Pass to different model/role
shard	1 → N	Split long content
verify	1 → 2	Trust contract (challenger + synthesizer)

LangGraph Tracer

from langchain_mmcp import MMCPTracer

tracer = MMCPTracer(regulation_tags=["SOC2", "GDPR"], export_path="./mmcp-audits/")

# Add to ANY LangGraph or LangChain pipeline
result = app.invoke(input, config={"callbacks": [tracer]})
tracer.print_summary()

Cloud Server

# Start locally
uvicorn mmcp_cloud.server:app --port 8765

# Or on Railway/Render with DATABASE_URL for PostgreSQL
DATABASE_URL=postgres://... uvicorn mmcp_cloud.server:app

API Endpoints:

• POST /v1/auth/register — Create account
• POST /v1/auth/login — Get API key
• POST /v1/chat/completions — Proxy with billing
• POST /v1/chat/completions/stream — SSE streaming
• POST /v1/billing/checkout — Stripe checkout
• GET /v1/account/usage — Usage stats

Environment

# Required (at least one)
export OPENROUTER_API_KEY=sk-or-...   # Recommended (multi-model)
export ANTHROPIC_API_KEY=sk-ant-...   # Direct Anthropic

# Optional (cloud server)
export DATABASE_URL=postgres://...     # PostgreSQL
export STRIPE_SECRET_KEY=sk_...        # Stripe billing

License

MIT