building-mcp-servers

Configures Celigo Tool and builder-mode APIs as MCP server endpoints for external AI agents.

3|2|Updated May 20, 2026
One-click install
npx skills add https://github.com/celigo/ai --skill building-mcp-servers-celigo
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: building-mcp-servers
Source: https://github.com/celigo/ai/tree/main/skills/building-mcp-servers
Command: npx skills add https://github.com/celigo/ai --skill building-mcp-servers-celigo

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you package Celigo Tool and builder-mode API capabilities into a Model Context Protocol (MCP) server so external AI agents can call your integrations as tools over a standardized endpoint.

Core Features & Use Cases

  • MCP server endpoint design: Create a unique relativeURI that identifies the MCP server path for clients.
  • Tool vs API exposure: Expose Celigo Tools (as MCP tools with optional annotations and overrides) and expose builder-mode APIs (as MCP tools without annotations/overrides).
  • Agent behavior hints: Configure MCP-standard annotations like readOnlyHint, destructiveHint, idempotentHint, and openWorldHint for safer, better tool usage.
  • Per-server customization: Use overrides (especially connection overrides) so the same underlying Tool can operate against different resources per server.
  • Validation guidance: Ensure tool input schemas are MCP-compatible (tool input must be an object at the root) and enforce name uniqueness across all exposed entries.

Quick Start

Use the building-mcp-servers skill to plan the capabilities you want to expose, then define an MCP server with a valid relativeURI and add tools[] and/or apis[] entries that reference existing Celigo resources.

Frequently Asked Questions about building-mcp-servers

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I expose Celigo tools to AI agents via an MCP server?

You expose Celigo tools by configuring an MCP server with a unique relativeURI and adding tools[] or apis[] entries that reference existing Celigo resources, making them callable endpoints for external AI agents.

What is the difference between exposing Celigo Tools and builder-mode APIs as MCP entries?

Exposed Celigo Tools support optional annotations and overrides for agent behavior hints and per-server customization, while builder-mode APIs are exposed as MCP tools without annotations or overrides.

Do I need a specific input schema format for tools exposed through a Model Context Protocol server?

Yes, tool inputs exposed through a Model Context Protocol server must use a root type of object to be MCP-compatible, and you must enforce name uniqueness across all exposed tool and API entries.

Can I use connection overrides to run the same Celigo Tool against different resources on separate MCP servers?

Yes, you can apply connection overrides to a tool so the same underlying Celigo Tool operates against different resources per MCP server, enabling per-server customization without duplicating the tool.

What MCP annotations should I configure to provide safer agent behavior hints?

You can configure MCP-standard annotations like readOnlyHint, destructiveHint, idempotentHint, and openWorldHint on exposed Celigo Tools to provide safer and better tool usage hints for agents.

Are there limitations on which Celigo APIs can be exposed as MCP tools?

Yes, you cannot expose script or legacy APIs as MCP entries; the configuration only supports exposing Celigo Tools and builder-mode APIs as callable endpoints for external AI agents.