The MCP interface provides discovery and execution of pre-built superglue
tools. It does not support ad-hoc system creation or tool building.
Use MCP in production for agentic use cases and internal GPTs.
When to use MCP
Controlled Access
Execute only pre-approved, tested tools. No ad-hoc code generation or system
modification.
Tool Execution in Production
Execute saved tools from your custom GPTs with full observability and error
handling.
Agent Workflows
Give AI agents the ability to discover and execute complex multi-system
workflows without manual API wiring.
Internal Tooling
Expose business logic as AI-accessible tools for customer support, ops
teams, or automated workflows.
Installation
Cursor
Add to your Cursor MCP settings (.cursor/config.json or via Settings → MCP):
Claude Desktop
Add to Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
Self-Hosted
Point to your self-hosted REST API and add a /mcp path:Available Tools
The MCP server exposes two core tools for discovery and execution:find_relevant_tools
Search for pre-built tools using natural language queries.
Input Schema:
- Discover available tools for a specific task
- Browse all tools (use
*as wildcard) - Find tools by system type or capability
execute_tool
Execute a saved tool by its ID.
Input Schema:
- Executes with saved credentials from systems used during build
- Full execution logs available in superglue dashboard
Usage Patterns
Discovery → Execution Flow
Troubleshooting
Connection Issues
Tool Not Found
find_relevant_tools to get valid tool IDs or check the dashboard.
Next Steps
Create a Tool
Build tools to expose via MCP
Creating a System
Connect and configure external systems
API Reference
Direct REST API access
Set Up Guide
Deploy self-hosted MCP server