mcp-hub

name	mcp-hub
description	MCP server integration hub — standardized tool access to databases, filesystems, memory, reasoning, and 100s of third-party services via Model Context Protocol. Use when querying databases, structured file ops, knowledge graph memory, step-by-step reasoning, or when the user asks about MCP servers.
triggers	["mcp server","query database","structured thinking","knowledge graph memory","mcp tool","add mcp server"]
category	meta
maturity	stable
tags	["mcp","model-context-protocol","mcporter","knowledge-graph","filesystem"]

MCP Hub — Model Context Protocol Integration

Central hub for all MCP server integrations. MCP provides standardized tool access — databases, files, APIs, and services through a common protocol.

Quick Reference

# List all configured servers
mcporter list

# List tools for a specific server
mcporter list <server> --schema

# Call a tool
mcporter call <server.tool> key=value

# Call with JSON args
mcporter call <server.tool> --args '{"key": "value"}'

Configured Servers

1. filesystem — File Operations

Controlled file access to project directories, workspace, skills, and /tmp.

# List directory
mcporter call filesystem.list_directory path=$HOME/projects

# Read a file
mcporter call filesystem.read_file path=${FOREX_FACTORY_SCRAPER_HOME:-$HOME/projects/forex-factory-scraper}/README.md

# Search files by name
mcporter call filesystem.search_files path=$HOME/projects pattern="*.py"

# Get file info
mcporter call filesystem.get_file_info path=${FOREX_FACTORY_SCRAPER_HOME:-$HOME/projects/forex-factory-scraper}/scraper.py

# Write a file
mcporter call filesystem.write_file path=/tmp/test.txt content="hello world"

# Move/rename
mcporter call filesystem.move_file source=/tmp/old.txt destination=/tmp/new.txt

Allowed directories: $HOME/projects, $HOME/.openclaw-poe/workspace, $HOME/openclaw/skills, /tmp

14 tools: read_file, read_text_file, read_multiple_files, write_file, edit_file, create_directory, list_directory, directory_tree, move_file, search_files, get_file_info, list_allowed_directories, read_image_file, tree

2. sqlite — Database Queries

Query SQLite databases. Default: kanban tasks DB.

# List tables
mcporter call sqlite.list_tables

# Describe a table
mcporter call sqlite.describe_table table_name=tasks

# Read query (SELECT only)
mcporter call sqlite.read_query query="SELECT * FROM tasks WHERE status='in_progress'"

# Write query (INSERT/UPDATE/DELETE)
mcporter call sqlite.write_query query="INSERT INTO tasks (title, status) VALUES ('test', 'todo')"

Default DB: ~/.kanban/tasks.db Switch DB: Edit ~/.mcporter/mcporter.json → change --db-path arg. Or use the switch script:

# Switch to meta-analyst DB
node ${CLAUDE_SKILLS_DIR:-$HOME/.claude-agent/.claude/skills}/mcp-hub/scripts/switch-db.cjs ~/.meta-analyst/events.db

# Switch to knowledge graph DB  
node ${CLAUDE_SKILLS_DIR:-$HOME/.claude-agent/.claude/skills}/mcp-hub/scripts/switch-db.cjs ~/.knowledge-graph/kg.db

# Switch back to kanban
node ${CLAUDE_SKILLS_DIR:-$HOME/.claude-agent/.claude/skills}/mcp-hub/scripts/switch-db.cjs ~/.kanban/tasks.db

6 tools: read_query, write_query, create_table, list_tables, describe_table, append_insight

3. memory — Knowledge Graph Memory

Persistent entity-relationship store. Stores facts as entities with observations and relations.

# Create entities
mcporter call memory.create_entities entities='[{"name":"the user","entityType":"person","observations":["Uses Telegram","Timezone Europe/Oslo"]}]'

# Create relations
mcporter call memory.create_relations relations='[{"from":"Poe","to":"the user","relationType":"assists"}]'

# Search by query
mcporter call memory.search_nodes query="forex"

# Open specific nodes
mcporter call memory.open_nodes names='["the user","Poe"]'

# Read entire graph
mcporter call memory.read_graph

# Add observations to existing entity
mcporter call memory.add_observations observations='[{"entityName":"the user","contents":["Prefers CET times over UTC"]}]'

# Delete entities
mcporter call memory.delete_entities entityNames='["test"]'

# Delete observations  
mcporter call memory.delete_observations deletions='[{"entityName":"test","observations":["old fact"]}]'

# Delete relations
mcporter call memory.delete_relations relations='[{"from":"A","to":"B","relationType":"knows"}]'

Storage: ~/.mcporter/memory-kg.json 9 tools: create_entities, create_relations, add_observations, delete_entities, delete_observations, delete_relations, read_graph, search_nodes, open_nodes

4. sequential-thinking — Structured Reasoning

Step-by-step problem decomposition for complex tasks.

mcporter call sequential-thinking.sequentialthinking \
  thought="First, let me identify the core problem..." \
  thoughtNumber=1 \
  totalThoughts=5 \
  nextThoughtNeeded=true

1 tool: sequentialthinking (with branching, revision support)

---

Adding New MCP Servers

Config location

~/.mcporter/mcporter.json

Adding a stdio server

{
  "mcpServers": {
    "my-server": {
      "command": "node",
      "args": ["/path/to/server.js"],
      "env": { "API_KEY": "..." },
      "description": "What this server does"
    }
  }
}

Adding a remote HTTP/SSE server

{
  "mcpServers": {
    "remote-server": {
      "url": "https://mcp.example.com/sse",
      "description": "Remote MCP server"
    }
  }
}

Testing

# Verify server starts
mcporter list my-server --schema

# Call a tool
mcporter call my-server.some_tool arg=value

---

What MCP Servers Can Be Used For

Categories of available servers (100s exist):

🗄️ Databases

• SQLite, PostgreSQL, MySQL, MongoDB, Redis, Supabase, Neon, PlanetScale
• Use: Query any database, inspect schemas, run analytics

📂 File Systems & Storage

• Local filesystem, Google Drive, Dropbox, S3, MinIO
• Use: Manage files across local and cloud storage

💬 Communication

• Slack, Discord, Telegram, Email (IMAP/SMTP), Microsoft Teams
• Use: Send messages, read channels, search conversations

🔄 Version Control

• GitHub, GitLab, Bitbucket (issues, PRs, repos, CI)
• Use: Manage repos, create issues, review PRs programmatically

📋 Project Management

• Linear, Jira, Notion, Trello, Asana, Monday.com
• Use: Create/update tickets, track sprints, manage backlogs

💰 Finance & Payments

• Stripe, Plaid, crypto APIs, stock market data
• Use: Monitor payments, check balances, analyze transactions

📊 Monitoring & Observability

• Sentry, Datadog, Grafana, PagerDuty
• Use: Check errors, view dashboards, acknowledge alerts

🔎 Search & Data

• Brave Search, Google Search, Tavily, web scraping, Wikipedia
• Use: Search the web, extract content, research topics

☁️ Cloud Platforms

• AWS, GCP, Azure, Vercel, Cloudflare, Docker
• Use: Manage infrastructure, deploy services, check status

🏠 Home Automation

• Home Assistant, MQTT, IoT device control
• Use: Control smart home devices, read sensors

🧠 AI & Knowledge

• Memory/knowledge graphs, RAG systems, vector DBs
• Use: Persistent memory, semantic search, context management

📧 Productivity

• Google Calendar, Todoist, Obsidian, Notion
• Use: Schedule events, manage tasks, organize notes

High-value servers to consider adding:

1. GitHub MCP — richer than gh CLI for bulk operations
2. Postgres MCP — if we add a Postgres DB for any project
3. Brave Search MCP — alternative to SearXNG
4. Sentry MCP — if we deploy production services
5. Docker MCP — manage containers programmatically
6. Notion/Linear MCP — if the user uses these for project management

---

Architecture Notes

• mcporter is the CLI bridge — it spawns stdio MCP servers on demand
• Config: ~/.mcporter/mcporter.json (system-wide) or ./config/mcporter.json (project)
• Each mcporter call spawns the server, calls the tool, gets the result, and exits
• For long-running use: mcporter daemon start keeps servers warm
• MCP servers are just programs that speak the MCP protocol over stdin/stdout or HTTP/SSE