Tool Box MCP Server
Tool Box MCP Server is a central AI tool registry with semantic/graph discovery and multi-step MCP tool chaining.
Discover tools: Search registered MCP servers, skills, tools, and commands by natural language using vector search.
Explore relationships: Expand tool dependencies or get complete tool clusters via the Knowledge Graph.
Manage registry: Register, list, and delete tools with metadata for MCP servers and skills.
Build and prepare chains: Auto-generate or manually define tool chains; analyze schemas, related skills, and recommended transforms before execution.
Execute chains: Run sequential MCP tool pipelines using
CHAIN_RESULT, JSONPath filtering, and built-in data transforms.Refresh chainable tools: Rediscover and list tools available for chaining from configured MCP servers.
Check health: Monitor ChromaDB, Knowledge Graph, registered tool count, and server version.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Tool Box MCP Server find tools for data analysis and excel export"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Tool Box MCP Server
Central registry for AI tools with Vector Search + Knowledge Graph for intelligent tool discovery and chaining.
Features
Vector Search: Semantic tool discovery via ChromaDB
Knowledge Graph: Tool relationships and dependency traversal
Tool Chaining: Execute multi-step MCP tool pipelines with data transformations
Progressive Disclosure: Load only what's needed
Unified Registry: MCP Servers, Skills, Tools, Commands in one place
Related MCP server: executor
Quick Start
# Clone
git clone https://github.com/Adriftnote/tool-box-mcp.git
cd tool-box-mcp
# Install dependencies
npm install
# Run setup (initializes ChromaDB and sample data)
./scripts/setup.sh
# Start server
npm startTools (11 total)
Tool | Description |
| Find tools by natural language query |
| Explore tool dependencies via Knowledge Graph |
| Get complete tool set for a task |
| Add new tools to registry |
| Remove tools from registry |
| List all registered tools |
| Auto-generate and execute chain from query |
| Execute sequential MCP tool pipeline |
| Analyze chain before execution (schemas, skills, transforms) |
| Refresh chainable tools from MCP servers |
| List tools available for chaining |
Installation
Prerequisites
Node.js 18+
Python 3.8+ (for ChromaDB)
pip (Python package manager)
Step 1: Clone and Install
git clone https://github.com/Adriftnote/tool-box-mcp.git
cd tool-box-mcp
npm installStep 2: Run Setup
# This installs Python dependencies and initializes ChromaDB
./scripts/setup.shStep 3: Configure MCP Client
Add to your Claude Code or MCP client settings:
{
"mcpServers": {
"tool-box": {
"type": "stdio",
"command": "node",
"args": ["/path/to/tool-box-mcp/dist/index.js"]
}
}
}Configuration
Data Files
After installation, configure these files in the data/ directory:
File | Purpose |
| MCP servers to chain with |
| Tool relationships |
| Vector search database |
Environment Variables (Optional)
Override default paths with environment variables:
TOOLHUB_GRAPH_PATH # Knowledge Graph JSON path
TOOLHUB_MCP_CONFIG # MCP config path
TOOLHUB_PYTHON_PATH # Python interpreter pathUsage Examples
1. Search Tools
toolhub_search({
query: "data analysis excel",
limit: 10,
include_graph: true
})2. Execute Tool Chain
toolhub_chain({
mcpPath: [
{
toolName: "sqlite_read_query",
toolArgs: "{\"query\": \"SELECT * FROM metrics LIMIT 10\"}",
outputTransform: "sqlite→2d"
},
{
toolName: "document_create_excel",
toolArgs: "{\"filepath\": \"/tmp/report.xlsx\", \"content\": \"CHAIN_RESULT\"}"
}
]
})3. Register New Tool
toolhub_register({
name: "my-mcp-server",
type: "MCP_Server",
description: "My custom MCP server for data processing"
})Architecture
┌─────────────────────────────────────────────┐
│ Claude Code / AI Agent │
└─────────────────────────────────────────────┘
│ MCP Call
▼
┌─────────────────────────────────────────────┐
│ Tool Box MCP Server │
│ ┌─────────────────────────────────────┐ │
│ │ HybridSearchService │ │
│ │ ├─ VectorSearchService (ChromaDB) │ │
│ │ └─ GraphSearchService (JSON) │ │
│ ├─────────────────────────────────────┤ │
│ │ ToolDiscoveryService │ │
│ │ └─ Chain Execution + Transforms │ │
│ └─────────────────────────────────────┘ │
└─────────────────────────────────────────────┘
│ │
▼ ▼
┌───────────────┐ ┌───────────────┐
│ ChromaDB │ │ Knowledge │
│ (data/) │ │ Graph JSON │
└───────────────┘ └───────────────┘Data Transforms
Built-in transforms for chain data flow:
Transform | Description |
| SQLite results to 2D array |
| Parse JSON string to object |
| Wrap object in array |
| Flatten nested arrays |
| Extract first element |
| Get object keys |
| Get object values |
Development
# Build from source
npm run build
# Run server
npm start
# Type check
npm run typecheckProject Structure
tool-box-mcp/
├── dist/ # Compiled JavaScript (included)
├── src/
│ ├── index.ts # Main server entry
│ ├── schemas/ # Zod input schemas
│ └── services/ # Search, chain, transform services
├── scripts/
│ ├── setup.sh # Initial setup script
│ └── register-tool.py # Tool registration utility
├── data/
│ ├── chromadb/ # Vector database
│ ├── knowledge-graph.json
│ └── mcp-config.json
└── package.jsonTroubleshooting
"ChromaDB not found"
pip install chromadb sentence-transformers"Python not found"
Set the Python path:
export TOOLHUB_PYTHON_PATH=/usr/bin/python3"No tools found"
Run setup to initialize sample data:
./scripts/setup.shLicense
ISC
Available Tools
4 toolstoolhub_clusterGet Tool ClusterARead-onlyIdempotent
Get a complete tool cluster for a task query.
This combines vector search and graph expansion to return a complete set of tools needed for a task. Includes primary tools (semantic matches) and their dependencies (graph expansion), plus usage context.
Args:
query (string): Natural language task description
include_context (boolean): Include usage patterns and examples (default: true)
response_format ('json' | 'markdown'): Output format
Returns: { "primary": [ { "name": "n8n-workflow-builder", "type": "MCP_Server", ... } ], "dependencies": [ { "name": "n8n-node-templates", "type": "Skill", ... } ], "context": { "usagePatterns": ["Use MCP servers for data operations", ...], "examples": ["Primary workflow: n8n-workflow-builder - ...", ...] }, "stats": { "totalTools": 7, "tokenEstimate": 7000, "savingsPercent": 92.1 } }
Use this tool for complete task setup - returns everything needed to start working.
Inputs: query (required), include_context, response_format.
toolhub_expandExpand Tool GraphARead-onlyIdempotent
Expand a tool to find its dependencies via Knowledge Graph.
Given a tool name, traverse the Knowledge Graph to find related tools, requirements, and dependencies. Useful for understanding what other tools are needed to complete a task.
Args:
tool_name (string): Name of the tool to expand (e.g., 'n8n-workflow-builder')
depth (number): Traversal depth (default: 2, max: 4)
relation_types (string[]): Filter by relation types (optional)
response_format ('json' | 'markdown'): Output format
Relation types:
REQUIRES: Tool A requires Tool B to function
WORKS_WITH: Tools that commonly work together
EXECUTABLE_VIA: Tool can be executed via another tool
OUTPUTS_TO: Tool outputs data to another tool
BENEFITS_FROM: Tool benefits from using another tool
Returns: { "tool": "n8n-workflow-builder", "dependencies": [ { "name": "n8n-node-templates", "type": "Skill", "relation": "BENEFITS_FROM" } ], "totalCount": 4 }
Use this tool when you know a primary tool and need to find related tools.
Inputs: tool_name (required), depth, relation_types, response_format.
toolhub_healthHealth CheckARead-onlyIdempotent
Check Tool Hub service status.
Returns:
chromadb: Connection status to vector database
knowledgeGraph: Knowledge graph file status
toolCount: Number of registered tools
version: Server version
Inputs: no named parameters.
toolhub_searchSearch ToolsARead-onlyIdempotent
Search for relevant tools using Vector Search + Knowledge Graph.
This tool finds the most relevant MCP servers, skills, and tools for a given query. It uses semantic similarity (ChromaDB) to find matches and optionally expands results using the Knowledge Graph to include dependencies.
Args:
query (string): Natural language query (e.g., "n8n workflow automation")
limit (number): Maximum tools to return (default: 10, max: 50)
include_graph (boolean): Expand with Knowledge Graph (default: true)
response_format ('json' | 'markdown'): Output format (default: 'json')
Returns: JSON format: { "results": [ { "name": "n8n-workflow-builder", "type": "MCP_Server", "description": "Create and manage n8n workflows", "similarity": 0.89 } ], "stats": { "vectorCount": 3, "graphCount": 4, "totalCount": 7, "tokenEstimate": 7000, "savingsPercent": 92.1 } }
Examples:
"n8n 워크플로우 자동화" → n8n-workflow-builder, n8n-node-templates, ...
"TikTok 데이터 분석" → sqlite_tiktok_analytics, pandas-excel, ...
"Excel 리포트 생성" → pandas-excel, 데이터-구조-파악, ...
Use this tool when you need to find which tools are relevant for a task.
Inputs: query (required), limit, include_graph, response_format.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 4 tools
toolhub_search and toolhub_cluster substantially overlap: both take a natural language query and return relevant tools with optional Knowledge Graph expansion, making it unclear when to prefer one. toolhub_expand and toolhub_health are distinct, but the boundary between search and cluster is not crisp.
All tools share the toolhub_ prefix and snake_case, but the suffixes are inconsistent in style: search and expand are verbs, health is a noun, and cluster is a noun/verb. A pattern like toolhub_get_cluster or toolhub_check_health would be clearer.
Four tools is well-scoped for a tool-discovery service: search, dependency expansion, full cluster assembly, and health checking. Each tool has a clear operational role, and the count is neither thin nor bloated.
The core discovery workflow (query, expand, cluster, health) is covered with no dead ends. Minor gaps exist: there is no direct list-all-tools or get-tool-detail operation, and cluster largely subsumes search, but agents can accomplish the main task.
Related MCP Connectors
AI agent registry — search, discover, register, and connect agents via MCP.
A registry of AI agent tools — MCP servers, APIs, CLIs, SDKs — kept current by automated ingestion.
Pay-per-use tool marketplace for AI agents. Search, price-check, and call APIs via MCP.
Find, compare, and audit software for AI agents. Scored registry of tools and MCP servers.
Related MCP Servers
AlicenseNot gradedqualityDmaintenanceMCP tool discovery for AI agents — find, compare, verify tools across 35+ registries.3MIT- AlicenseNot gradedqualityBmaintenanceAn integration layer for AI agents that provides a catalog of tools from various sources (OpenAPI, GraphQL, MCP, etc.) and can be used as an MCP server for compatible agents.4,451MIT
- AlicenseNot gradedqualityBmaintenanceMCP server registry and marketplace that enables dynamic discovery, installation, and activation of tools at runtime, allowing AI agents to use new tools without restarting.129 npm7MIT
- AlicenseNot gradedqualityDmaintenanceFederating gateway for AI agents to discover and call tools from multiple MCP servers with intelligent search and dynamic tool registration.39MIT