---
title: "MCP Tools: The API for AI Agents"
description: "How Tedix turns any REST API into MCP tools that AI agents can discover, understand, and use — no code required."
published: 2026-04-30T17:14:01.790Z
updated: 2026-04-30T17:14:01.790Z
canonical: https://blog.tedix.dev/posts/mcp-tools-the-api-for-ai-agents
---
The Model Context Protocol (MCP) is becoming the standard way AI agents interact with the world. At Tedix, we've built a platform that lets any organization expose their APIs as MCP tools — config-driven, no handler code needed.

## From REST to MCP in Minutes

Every MCP app on Tedix is a set of tool configurations stored in D1. Each tool maps to an API endpoint with input/output schemas, auth requirements, and widget layouts. When an AI agent connects to your MCP server, it discovers tools dynamically — no code generation, no SDK updates.

## Progressive Skill Disclosure

AI agents don't need to see all tools at once. Tedix uses progressive disclosure — skills are summarized in the MCP server instructions, and agents load full procedures only when relevant. This keeps context windows lean and agents focused.

## Code Mode: One Tool to Rule Them All

For apps with dozens of tools, Code Mode collapses everything into a single 'code' tool. The AI agent writes JavaScript that calls namespaced functions — no tool selection overhead, full composability, and the ability to chain multiple operations in one turn.

## Three Transport Types

Tedix supports three ways to connect tools to backends: RPC (internal oRPC via service binding), MCP (upstream MCP server proxy), and External (third-party REST APIs with credential injection). Each transport type handles auth, retries, and error mapping transparently.

## Get Started

Create an app on the Tedix dashboard, add your API endpoints as tools, and connect any MCP client. Your tedi can even build and iterate on the app configuration autonomously — that's the Tedix flywheel in action.