Notes on building the agent runtime
Design decisions, failures, and lessons from building AI agents that live inside the messenger. Every product change lands in the changelog.
3Posts published
4Releases
2024Shipping since
Latest posts
View allDesigning an agent runtime that lives inside the messenger
vooy's agents run inside the messenger your team already uses. Here are the decisions that shaped how we folded sessions, tool calls, and streaming into one runtime.
4 min read#Runtime
Wiring up hundreds of tools with MCP: the connector-hub design
An agent is only useful if it can touch external tools. Here's the structure of connector-hub — how vooy standardizes dozens of SaaS integrations behind MCP — and the traps we hit along the way.
3 min read#Connectors
Streaming LLM responses reliably on Cloudflare Workers
vooy runs its agents, landing, and API entirely on Cloudflare. Here's how we stream long LLM responses without stalls at the edge — worker lifetimes, backpressure, and the OpenNext story.
3 min read#Cloudflare