I've just been using Deep Research on ChatGPT and Perplexity a lot to write and research complex technical reports for my boss. He loves the reports and it has decreased my workload a ton but I still have some frustrations with it. None of them provide an API that gets me the same quality of output you would with the applications.
I wanted something with more control on the LLMs, swappable with the reasoning new models that came out.
Not just prompt → response; I wanted to shape the way the reasoning happened. How prompting was made, the design of the process, how many layers of thinking the models should go through, and what information the LLM should be extracting from the internet.
Essentially I wanted the following things: • Choose which LLM models were used and how they were combined • Adjust how deep or broad the exploration went • Specify the report format: token count, structure, even tone • Expose verified sources that were used throughout the process
But most of the deep research tools I’ve used don't allow that or has some weird edge case. They are mostly closed system/sources or commercial API’s for end-users - NOT developers.
So I built Deep Research - a fully customizable open source deep research framework. It’s designed for devs who want full control over the research process, without building everything from scratch.
It supports: • Multi-model orchestration — Plug in your own LLMs or use the defaults (GPT-4.1 is baked in) • Custom depth and breadth — Tune the research from shallow summaries to multi-layered deep dives • Token-level control — Define how long the output should be • Transparent citations — Everything is traceable back to its source • Built-in web search — Integrates JigsawStack’s web search to combine retrieval with reasoning
It runs on top of Vercel’s AI SDK making it easily configurable!
Here’s a full blog breakdown: https://jigsawstack.com/blog/introducing-jigsawstack-deep-re...
Try it out. Would love to hear feedback from fellow hackers! What’s good, what’s bad, what’s ALARMING.