JSON Editing with LLMs - Getting it right
We went through three architectural iterations to figure out how our AI agent should apply edits to JSON tool configurations. Here's what we learned about structured output, search & replace, and JSON Patch.
Latest thoughts from the team
Insights on API integrations, workflow automation, and building reliable agent systems for enterprise.
Why Fetching Documentation for Context-Engineering is Hard (and What We Built at superglue)
In building integrations using AI, one of the toughest engineering tasks is turning third-party documentation into reliable context for agents. Learn about our approach at superglue.
A New Imperative for Developer-facing UIs
Why developer tools don't need UIs anymore. With MCP and agents like Cursor, the age of clicking through dashboards is ending. The future is capability-based interfaces that work directly in your editor.
Why do we even need MCP?
The fundamental difference between agentic tools and API endpoints. Why MCP exists as an abstraction layer that provides capabilities rather than forcing agents to navigate endpoint orchestration.
What most people get wrong about MCP
The reality behind MCP adoption: who builds and maintains MCP servers for legacy enterprise systems? Exploring the challenges of bridging modern AI interfaces with decades-old software.