/connect

MCP COORDINATION NETWORK
========================

Coordinator Instance (Hub)
    |
    +-- Blackboard (Shared State via ETS/Horde)
    |   |
    |   +-- Active Tasks Registry
    |   +-- Instance Registry (heartbeat-based)
    |   +-- Sync State (vector clocks)
    |   +-- Message Queue (ordered delivery)
    |
    +-- Worker Instance 1 (Spoke)
    |   +-- Local State Cache
    |   +-- Task Queue (priority-ordered)
    |   +-- Health Reporter
    |
    +-- Worker Instance 2 (Spoke)
    |   +-- Local State Cache
    |   +-- Task Queue (priority-ordered)
    |   +-- Health Reporter
    |
    +-- Observer Instance (Read-Only)
        +-- State Mirror (eventual consistency)
        +-- Metrics Collector

# Check coordination status across all servers
/connect status

# List all active instances with health information
/connect list

# Join coordination network as a worker
/connect join --role worker

# Join as the coordinator (only one per network)
/connect join --role coordinator

# Join as read-only observer
/connect join --role observer

# Broadcast a message to all connected instances
/connect broadcast "Starting integration phase"

# Synchronize state with all instances
/connect sync

# Leave coordination network gracefully
/connect leave

# Force reconnection to specific server
/connect reconnect prismatic-mcp

# Check health of specific server
/connect health postgres-server

# Restart all MCP server connections
/connect restart-all

# Show detailed connection metrics
/connect status --verbose

/connect [action] [options]
    |
    v
PHASE 1: VALIDATION (< 100ms)
    +-- Verify action parameter
    +-- Check authority level (L2+)
    +-- Validate instance_id if provided
    +-- Load server configuration
    |
    v
PHASE 2: DISCOVERY (< 500ms)
    +-- Enumerate configured MCP servers
    +-- Probe server availability (parallel health checks)
    +-- Build connection topology map
    +-- Identify coordinator instance
    |
    v
PHASE 3: EXECUTION (< 2s)
    +-- Execute requested action
    +-- Update blackboard state
    +-- Propagate state changes to peers
    +-- Confirm action completion
    |
    v
PHASE 4: REPORTING
    +-- Display action result
    +-- Show updated connection status
    +-- Log telemetry events
    +-- Update instance registry

MULTI-INSTANCE COORDINATION STATUS
==================================

This Instance:
  ID: claude-instance-a7b3
  Role: worker
  Status: connected
  Uptime: 2h 14m

Connected Instances: 3
  - claude-instance-f2c1 (coordinator) [active]
  - claude-instance-a7b3 (worker) [active] <- YOU
  - claude-instance-e9d4 (worker) [active]

MCP Servers: 14/14 healthy
  - prismatic-mcp .......... [OK] 27 tools
  - filesystem ............. [OK] 8 tools
  - github ................. [OK] 12 tools
  - postgres ............... [OK] 6 tools
  - memory ................. [OK] 4 tools
  - context7 ............... [OK] 3 tools

Blackboard:
  Active Tasks: 5
  Pending Sync: 0
  Last Sync: 2 seconds ago

Health: EXCELLENT

# Initialize all MCP servers
./scripts/setup-mcp-servers.sh

# Verify server configuration
/connect status --verbose

# If a server becomes unresponsive
/connect reconnect <server-name>

# If coordination state is corrupted
/connect sync --force

# If all servers need restart
/connect restart-all --staggered

# Terminal 1: Launch as coordinator
./scripts/dev/launch-multi-claude.sh
/connect join --role coordinator

# Terminal 2: Launch as worker
/connect join --role worker

# Terminal 3: Launch as observer for monitoring
/connect join --role observer

# Coordinator distributes tasks
/connect broadcast "Phase 1: Parallel test execution"

# Check connection status programmatically
{:ok, status} = PrismaticClaude.MCP.connection_status()

# Register a new MCP server at runtime
:ok = PrismaticClaude.MCP.register_server(%{
  name: "custom-server",
  url: "http://localhost:8080",
  tools: ["custom_tool_1", "custom_tool_2"]
})

# Monitor connection health via telemetry
:telemetry.attach("mcp-monitor",
  [:prismatic_claude, :mcp, :connection],
  &handle_connection_event/4,
  nil
)

{
  "prismatic-mcp": {
    "command": "node",
    "args": ["dist/index.js"],
    "env": {
      "PRISMATIC_API_KEY": "${PRISMATIC_API_KEY}"
    },
    "priority": 1,
    "health_check_interval": 30000,
    "restart_policy": "always"
  }
}

Related Commands