Notes on putting your app inside AI agents.

What we learn building Gephyr: MCP, agent-ready APIs and keeping write access safe.

1. 20 October 2026

   Agent-Ready APIs: 8 Things That Break AI Agents (and How to Fix Them)

   We connected three public APIs to real AI agents through MCP and watched what broke. The eight gaps that trip agents up, how to fix each one, and what Gephyr handles for you.

   Read →
2. 20 October 2026

   MCP Security: What Could Go Wrong When AI Agents Can Change Things

   Prompt injection, broken authorization, SSRF, runaway agents and leaked credentials: the real MCP security risks of AI agent write access, and the controls that limit the damage.

   Read →
3. 20 October 2026

   Stop Building AI Into Your App. Let the AI Come to You.

   MCP vs embedded AI: why exposing your app to ChatGPT, Claude and other AI agents through MCP beats building another in-app chatbot.

   Read →