Cybrid MCP Server

Cybrid's Model Context Protocol (MCP) server gives AI coding assistants accurate Cybrid API context—including endpoints, request bodies, response schemas, and ready-to-use code snippets—inside your editor.

Server URL: https://docs.cybrid.xyz/mcp

What is MCP?

Model Context Protocol (MCP) is an open standard that allows AI applications to securely access external data sources and tools. The Cybrid MCP server provides AI agents with

Documentation search capabilities
Code generation assistance for Cybrid integrations

What You Can Do with Cybrid MCP

Once connected, your IDE's AI agent can

Search Cybrid specs and docs
List endpoints and security schemes
Pull request/response schemas and example bodies
Generate code snippets for specific endpoints

⚠️
Security Limitation
The MCP server does not execute API calls. AI tools can only access API documentation and schemas, not invoke Cybrid endpoints.

Install in Popular IDEs

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "cybrid": {
      "url": "https://docs.cybrid.xyz/mcp"
    }
  }
}

Available Tools

The Cybrid MCP server exposes these MCP tools to clients. Tools are organized by workflow stage: Discovery (finding endpoints), Reference (getting details), Schema (request/response contracts), Snippets (code generation), Auth (security schemes), Environments (server configuration), and Docs (documentation search).

⚠️
Security Limitation
The execute-request tool is intentionally disabled to prevent AI tools from automatically invoking Cybrid API calls.

Tool	Category	Description	When to use
list-specs	Discovery	List available Cybrid OpenAPI specs that can be queried.	Use when you want to know which API specs are available (e.g., sandbox vs production).
search-specs	Discovery	Search across specs by keyword.	Use to find relevant endpoints by topic (e.g., "transfers", "quotes", "webhooks").
list-endpoints	Discovery	List endpoints, optionally scoped by spec or tag/category.	Use to enumerate all endpoints for a feature area.
get-endpoint	Reference	Get full details for a specific endpoint (method, path, description, params).	Use after you've found an endpoint and want the exact contract.
get-request-body	Schema	Retrieve the request body schema (and/or example) for an endpoint.	Use to generate strongly-typed request objects or payload examples.
get-response-schema	Schema	Retrieve the response schema for an endpoint.	Use to validate parsing logic and generate types/interfaces.

General Usage with MCP Servers in AI Chat Assistants

When you connect an MCP server to an AI chat assistant (Cursor, VS Code Copilot Chat, Windsurf, Claude, etc.), the assistant can call MCP tools to pull authoritative, structured context (endpoints, schemas, examples) into the conversation.

A few tips that consistently work well:

Be specific about what you want back. Ask for endpoint details, request bodies, response schemas, or snippets explicitly.

Name the MCP server if you have more than one connected (e.g., "Use the cybrid MCP server…").

Start broad, then narrow. Use search/list tools first, then fetch the exact endpoint/schema once you've found the right match.

Ask for outputs in the format you'll use. For example: "Return the request body as JSON + a TypeScript interface."

Remember the safety model: Cybrid MCP does not allow auto-executing API calls (execute-request is disabled). It can help you build correct requests, but it won't run them.

Example 1: Find an Endpoint by Topic, Then Pull the Request Body

Prompt:

"Using the cybrid MCP server, search the specs for quote creation. Then show me the endpoint and its request body."

What the assistant should do:

Call search-specs → identify relevant spec/paths
Call get-endpoint → confirm method/path + params
Call get-request-body → return schema/example payload

Example 2: Generate a Code Snippet for a Specific Endpoint

Prompt:

"Use cybrid: get me a TypeScript code snippet for the endpoint that creates a transfer. Include required headers."

What the assistant should do:

Call search-specs or list-endpoints → locate "create transfer"
Call get-endpoint → confirm exact endpoint
Call get-code-snippet → generate TS snippet