Building MCP servers using Gram from Speakeasy скачать в хорошем качестве

Building MCP servers using Gram from Speakeasy

In this demo, Georges Haidar, software engineer at Speakeasy, walks us through Gram, a new platform for creating AI agent tools and MCP servers from APIs. Watch as he transforms the massive Stripe API (600+ endpoints) into focused, usable tool sets that AI agents can actually work with effectively. Learn about the sweet spot for tool count and best practices, how to improve tool descriptions for better agent understanding, and see a powerful demo of cross-API workflows that combine HubSpot, Stripe, and Speakeasy APIs into a single business process solution. Perfect for developers looking to build practical AI agents that can handle real-world business tasks. Check out Gram at https://www.getgram.ai/ Watch Georges's developer meetup talk:    • Bridging the gap between your API and LLMs   Watch the interview preceding this demo:    • How Speakeasy approaches API design for AI...   chapters 00:00 Introduction to gram and MCP servers 01:36 Creating tool sets from stripe API 03:03 Context window limitations with too many tools 04:23 Optimal tool count for AI models 05:51 Improving tool descriptions and names 07:25 API rate limiting and proxy architecture 08:49 Credential management and environments 10:43 Multi-API tool sets and business processes 13:02 Custom workflow tools and structured prompts 14:58 Live demo of cross-API business workflow 16:05 Future iterations and closing thoughts