Building a Deep Research Agent Using MCP-Agent

2 days ago (thealliance.ai)

I gotta say, having white blurry blobs of something in the background floating behind white/grey text maybe wasn't the best design-choice out there.

None the less, I tried to find the actual APIs/service/software used for the "search" part, as I've found that to be the hardest to actually get right (at least for as-local-as-possible usage) for my own "Deep Research Agent".

I've experimented with Brave's search API which worked OK, but seems pricey for agent usage. Currently experimenting with using my own (local) YaCy instance right now, which actually gives me higher quality artifacts at the end, as there are no rate-limits and the model can do hundreds of search calls without me worrying about the cost. But it isn't very quick at picking up some stuff like news and more, otherwise works OK too.

What is the author doing here for the actual searching? Anyone else have any other ideas/approaches to this?

Great write-up! Gives me a few ideas for a governance bot that I'm working on. Thanks for sharing :)

A good model for planner seems pretty important, what models are best?

  • OP here -- I think the general principle I would recommend is using a big reasoning model for the planning phase. I think Claude Code and other agents do the same. The reason this is important is because the quality of the plan really affects the final result, and error rates will compound if the plan isn't good.

  • based on the article, it seems like a good reasoning model like gpt5 or opus 4.1 might be good choices for the planner. I wonder if the gpt oss reasoning models would do well

    • Personally been using GPT-OSS-120b locally with reasoning_effort set to `high` and it blows pretty much every other local model out of the water, but takes a lot of time for it to eventually do a proper content reply. But for fire-and-forget jobs like "Create a well-researched report on X from perspective Y" it works really well.

      1 reply →