I have been trying to integrate my life with AI for the last 3-4 years. When LLMs came out, I built a few AI-powered tools to query my health and personal data. I realized Agents were incurring a huge discovery cost trying to figure out which API to call, what the data schema would look like, etc. So I built the Just-In-Time API (JITAPI). An MCP server that helps the LLM figure out API endpoints at runtime without overloading the context with huge API specs. Once the harnesses arrived, I started integrating different parts of my life, such as health, Money, and home, by writing custom code and sharing credentials with the agents. The idea of agents having access to everything while I am flying blind about what they have access to sounded scary, so I built Clawlense. An OpenClaw plugin that shows you exactly what the agents are up to. While doing all of this, I noticed a pattern 1. Every agent paid a discovery cost, trying to figure out what API to call and what the data schema would look like 2. Rewriting of the glue code for different agent harnesses and platforms. 3. How insecure it is to share credentials and live tokens with agents, and doing it the right way is complex. 4. No visibility into what an agent has access to, no granular control over their access. Postern solves all four of the above and more. It has been running my life for the last 2 months. 14 connections, 8 services, 4 agents on top.

Postern

One MCP gateway to your money, health, mail, and home

One MCP gateway to your money, health, mail, and home