spt-mcp-server

SmartPoints MCP Server --- trust operations for Copilot and agent integrations

Overview

spt-mcp-server is an HTTP-based MCP server (binary crate) that exposes SPT trust and identity operations as tools for Copilot, Claude, and other AI agent integrations. It provides REST endpoints for agent identity verification, container validation, compliance checking, SLA negotiation, and trust graph queries. Built on Axum with BLAKE3-based content hashing.

Key Features

• Agent Identity Verification --- verify and create agent DIDs with trust scoring
• Container Validation --- validate signed SPT containers against a registry
• Compliance Checking --- evaluate agents against configurable policy rules
• SLA Negotiation --- negotiate service-level agreements between agents
• Trust Graph Queries --- query the agent trust graph for relationships
• Audit Trail --- retrieve tamper-evident audit logs
• Ephemeral Memory Status --- inspect short-lived memory allocations

Dependencies

Key external dependencies:

• axum --- HTTP framework
• tower-http --- CORS and tracing middleware
• blake3 --- content hashing
• uuid --- agent identity generation

No internal SPT crate dependencies (standalone server).

Usage

# Run the server
cargo run --bin spt-mcp-server
# Server listens on the configured port (default from env)

Endpoints

Method	Path	Description
GET	/health	Health check
POST	/mcp/tools/list	List available MCP tools
POST	/mcp/tools/call	Invoke an MCP tool by name

API Reference

MCP Tools

Tool	Description
verify_agent_identity	Verify an agent's DID and trust tier
create_agent_identity	Register a new agent identity
validate_spt_container	Validate a signed container hash
check_compliance	Run compliance policy checks
negotiate_sla	Negotiate an SLA between agents
query_trust_graph	Query trust relationships
audit_trail	Retrieve audit log entries
ephemeral_memory_status	Check ephemeral memory state

Core Types

Type	Description
McpToolCall	Incoming tool call request (name + arguments)
McpToolResult	Tool result with content blocks and error flag
AppState	Shared state: agents, containers, policies
AgentRecord	Agent identity with DID, trust tier, credentials
ContainerRecord	Signed container metadata
ComplianceCheck	Policy requirement check result