What is MCP?
The Model Context Protocol (MCP) is an open standard for delivering structured context to AI models and agents. It defines how context providers communicate with AI systems — specifying message formats, quality metadata, and delivery semantics.
mantle is MCP-native from the ground up. Every context response is delivered as a standards-compliant MCP message, ensuring compatibility with any MCP-enabled agent framework.
How mantle implements MCP
When an AI agent queries mantle, the response is structured as an MCP context message containing quality-scored objects, source provenance, and relationship metadata. Here is an example response:
{
"protocol": "mcp/1.0",
"type": "context_response",
"query": "Q3 revenue by region",
"context": {
"objects": [
{
"id": "obj_a1b2c3",
"type": "metric",
"content": "Q3 North America revenue: $42.3M",
"source": "snowflake",
"quality_score": 0.96,
"freshness": "2026-03-29T10:30:00Z"
}
],
"relationships": [...],
"total_objects": 12,
"sources_queried": 4
},
"metadata": {
"quality_score": 0.94,
"latency_ms": 87,
"token_count": 1847
}
}Supported Capabilities
Context Search
Natural language queries across all connected sources with quality scoring and relevance ranking.
Entity Resolution
Cross-source entity matching and relationship discovery via the semantic knowledge graph.
Streaming Responses
Server-sent events for real-time context delivery as objects are resolved across sources.
Quality Metadata
Every object includes quality score, freshness timestamp, source provenance, and confidence level.
Token Optimization
Context payloads are optimized for token efficiency, reducing costs when passed to language models.
Audit Trail
Full provenance tracking — every data point carries its origin, transformation history, and access permissions.