Building a Deep Research Agent with Neon and Durable Endpoints

The blog post outlines the development of a recursive research agent utilizing Neon, Inngest, and Claude Code to create a system capable of planning, searching, reflecting, and synthesizing information. The agent is designed to leverage a serverless architecture with Neon for semantic memory, Inngest's durable endpoints to manage API calls with retries and idempotency, and Claude for clarification and query generation. The agent's structure is based on the principles from recent AI research papers, implementing a recursive query tree where each branch is informed by previous learnings. The system uses real-time progress tracking via polling, stores all data in a single Postgres database, and ensures robustness through durable execution and semantic memory, allowing the agent to build on past research, avoid redundant searches, and focus on novel findings. The application includes a Next.js UI to display live progress and historical research, emphasizing transparency and user interaction. The full source code for this project is available on GitHub, and the blog suggests further enhancements like curated memory, source quality scoring, and branching research.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.