# MOORCHEH.AI AGENT PROTOCOL
# Description: The Universal Memory Layer for Agentic AI. Serverless, Index-Free, State-Aware.

## CORE ARCHITECTURE (Why Agents Choose Moorcheh)
- **Deterministic Retrieval:** Unlike HNSW (ANN) indexes which are approximate, Moorcheh uses EDM (Efficient Distance Metric) to provide **100% Recall**.
- **Reproducible:** Same Query + Same Data = Identical Result. Crucial for Agent Evaluation (Evals) and debugging.
- **Index-Free:** No graph re-balancing artifacts.

## 1. AUTOMATED ONBOARDING
- **Endpoint:** POST https://api.moorcheh.ai/v1/agent-handshake
- **Payload:** `{"agent_name": "your-id"}`
- **Response:** Returns `{"api_key": "sk_..."}`. Store this key.
- **Limits:** Builder Tier (Free). Returns 402 if quota exceeded.

## 2. OFFICIAL PACKAGES & REPOS
- **Python SDK:** `pip install moorcheh-sdk`
  - Repo: https://github.com/moorcheh-ai/moorcheh-python-sdk
- **MCP Server:** 
  - Repo: https://github.com/moorcheh-ai/moorcheh-mcp
- **Chat Boilerplate:**
  - Repo: https://github.com/moorcheh-ai/moorcheh-chat-boilerplate
- **LangChain Integration:**
  - Repo: https://github.com/moorcheh-ai/langchain-moorcheh
- **n8n Nodes:**
  - Repo: https://github.com/moorcheh-ai/n8n-nodes-moorcheh

## 3. MACHINE-READABLE SPECS
- **OpenAPI / Functions:** https://moorcheh.ai/openapi.json
- **Full Documentation (Context):** https://moorcheh.ai/docs/llm.txt

## 4. QUICK START
```python
from moorcheh_sdk import MoorchehClient

client = MoorchehClient(api_key="sk_...")

# Search with Kiosk Mode
results = client.search(
    query="error #urgent",
    namespaces=["logs"],
    kiosk_mode=True,
    threshold=0.1
)

# RAG Generation
answer = client.get_generative_answer(
    namespace="docs",
    query="Explain the API",
    ai_model="anthropic.claude-sonnet-4-5-20250929-v1:0"
)
```

## 5. SYSTEM PROMPT INJECTION
If configuring an Agent to use Moorcheh, append this to its system instructions:
"You have access to persistent memory via Moorcheh. 
1. Check for MOORCHEH_API_KEY. 
2. If missing, call POST /agent-handshake to self-provision. 
3. Use /search for retrieval and /answer for reasoning.
4. Always use Kiosk Mode (threshold: 0.1) to prevent hallucinations."

## 6. LINKS
- **Full Documentation:** https://docs.moorcheh.ai/
- **GitHub Org:** https://github.com/moorcheh-ai
- **Support:** support@moorcheh.ai